American Diabetes Association - 72nd Scientific Sessions

June 8-12, 2012; Philadelphia, PA; Report - Artificial Pancreas, CGM, Pumps, SMBG - Draft

Executive Highlights

ADA gave us encouraging updates on all things diabetes technology. Below, we enclose our broad takeaways and themes from the four major topic areas in this report: CGM, SMBG, insulin pumps, and the artificial pancreas (AP). As expected, the AP really stole the show at ADA, with particular focus on making systems portable, the newest products from Medtronic and Animas, infusing both insulin and glucagon, and debate over the biggest obstacles that remain.

  • ADA 2012 was full of insight on the next generation of CGM technology. Dr. David Price (Dexcom, San Diego, CA) shared new accuracy data from Dexcom’s pivotal study of the G4 sensor: a mean absolute relative difference (MARD) of 13%, 80% of points in the CEG A-Zone, and 94% of sensors lasting up to seven days (1-OR). As a reminder, the G4 was submitted to FDA at the end of 1Q12 and management hopes for approval before year-end (see Dexcom 1Q12 at http://www.closeconcerns.com/knowledgebase/r/109d0417). In the same session, Dr. Steven Russell (Massachusetts General Hospital, Boston, MA) unveiled data from the team’s closed-loop experiments comparing the Dexcom G4, the Medtronic Enlite, and the FreeStyle Navigator (as an aside, it’s great to see independent comparative accuracy data!). The G4 was the most accurate (MARD of 11.3%), although we were also glad to see the improvements in the Enlite sensor relative to the Sof-Sensor (MARD of 16.0- 17.2% vs. 20.3% for the Sof-Sensor) (4-OR). New looks at next-generation CGM technology also included 24-hour pilot study data from Echo Therapeutics’ transdermal CGM (MARD of 12.6%) (7-OR) and a poster summarizing the pivotal six-day trial of Medtronic’s Enlite CGM sensor (depending on the calibration scheme, a MARD of 13.6-14.7% and CEG A-zone 81-86%) (30-LB). As a reminder, Enlite was recently submitted to FDA along with the MiniMed 530G low glucose suspend pump; approval is expected in 1H13 (see our report at http://www.closeconcerns.com/knowledgebase/r/cbcebe6a). [Editor’s note: CGM accuracy data noted above may not be comparable due to differences in study design, calibration schemes, etc.]
  • We also noticed broader and more refined thinking about CGM – and a lot more use of CGM in clinical trials. Behavioral factors came up much more often in presentations and Q&A, especially from perplexed questioners wondering why so many people quit CGM and how they might be encouraged to stay on the technology. We appreciated this heightened focus on behavioral and psychological factors and believe it will: (1) take the field to a higher level in terms of product development; (2) help fine-tune patient selection criteria and education; and (3) help put (and keep) more patients on the technology, and ultimately get to the point where CGM represents the new standard of care. Also noteworthy were two excellent presentations on interpreting CGM downloads from Drs. Bruce Bode (Emory University, Atlanta, GA) and Howard Wolpert (Joslin Diabetes Center, Boston, MA) – we believe this is another area with significant potential to move the needle on CGM adoption at both a micro level (e.g., improving individual patients’ glycemic control through better retrospective analysis) and a macro level (e.g., boosting the utility of CGM for HCPs). The broader level of thinking about CGM also featured illuminating talks on CGM cost and coverage issues (Dr. Michael O’Grady), glycemic variability (Drs. David Rodbard, Robert Vigersky, Han DeVries, James Krinsley, and Thomas Danne), and how study design and analytic techniques affect reported CGM accuracy (Dr. David Price). We also came away from this year’s ADA with the sense that CGM is becoming increasingly used in clinical trials for drugs. This extra data should offer a plethora of additional insight over just A1c or SMBG profiles, especially to differentiate basal insulins, GLP-1 receptor agonists, and ultra-rapid acting insulins. Just as when GLP-1 came into the fold and suddenly everyone was asking about, with novel new drugs, “What was the change in weight?”, we feel like we are getting to the point where when novel drug results are presented, it is increasingly being asked, “Did you use CGM?” or, “What happened with CGM?”
  • Highlights on SMBG were mostly in the exhibit hall, with a major focus on better software. The three meters “launched” at ADA – Sanofi’s iBGStar, Abbott’s FreeStyle InsuLinx, and LifeScan’s OneTouch Verio IQ – were all heavily marketed with easier data review and analysis in mind (for our thoughts on the three meters, see Closer Look reports at http://www.closeconcerns.com/knowledgebase/r/afddbe77 and http://www.closeconcerns.com/knowledgebase/r/07b9a2d3 and diaTribe reviews at http://www.diatribe.us/issues/43/test-drive and http://www.diatribe.us/issues/41/test-drive). SMBG software has been suboptimal for the longest time, so this next- generation of easier-to-download meters, more actionable data reports, and on-meter statistics and pattern recognition is very refreshing to see from a patient perspective. In their corporate- sponsored events, Sanofi, Abbott, and LifeScan did a particularly good job attracting some of the field’s notable speakers to advocate for their products: Drs. Bruce Bode (Emory University, Atlanta, GA), Steve Edelman (UCSD, San Diego, CA), Bill Polonsky (Behavioral Diabetes Institute, San Diego, CA), and Ralph DeFronzo (University of Texas Health Science Center, San Antonio, TX). Bayer’s exhibit hall booth showcased the recently approved Contour Next EZ and Next Link, which were introduced at ADA in anticipation of the big upcoming US launch in late summer or early fall. The strips have achieved a very impressive level of accuracy: 99% of the time within 10% of lab standards (>75 mg/dl) and 100% within 10 mg/dl (<75 mg/dl). This is a major achievement in our view and was some of the very biggest news at the exhibit hall overall at ADA Presentations on SMBG were primarily in a joint ADA/EASD symposium, highlighted by a debate on the utility of SMBG for type 2s on oral agents (more data needed, though structured testing is important) and a talk on accuracy standards (the upcoming new ISO 15197 requirements are a move in the right direction [95% within 15 mg/dl (<100 mg/dl) or 15% (>100 mg/dl)]).
  • Valued information on insulin pumps came from the exhibit hall, headlined by the launch of Tandem’s t:slim. Tandem held a well-attended product theater to introduce its new touchscreen pump, which included presentations from CEO Kim Blickenstaff, expert educator Ms. Jen Block (Stanford University, Stanford, CA), and Dr. Timothy Bailey (Advanced Metabolic Care and Research, Escondido, CA). The company began taking orders on June 11 and will begin shipping the t:slim in August (we learned at ADA that the new pump will also ship with the LifeScan OneTouch Verio IQ blood glucose meter). Although Insulet did not secure FDA approval for the second-generation OmniPod in time for ADA, the new pod was on display in the exhibit hall accompanied by ample marketing materials and signage. As of Insulet’s 1Q12 results call in early May (see our report at http://www.closeconcerns.com/knowledgebase/r/d7d2d29f), approval for the next-gen pod is expected in the coming months. The center of Medtronic’s exhibit hall booth was devoted to the mySentry remote monitor (see our report on the approval at http://www.closeconcerns.com/knowledgebase/r/aca26119), the only product of its kind right now in the pump/CGM arena. Despite its $3,000 price tag, this device is starting to gain support from patients (~1,000 users right now) and even some small payer support. In the expansive J&J LifeScan/Animas booth, the Animas Vibe was on display, though reps were quick to note that the device has not yet been submitted for FDA review (as of Dexcom’s 1Q12 call, this is on hold for the moment). Roche had a kiosk devoted to its Accu-Chek insulin delivery products, while Cellnovotold us it hopes to launch in the US in 2013. Finally, Valeritas’ V-Go (recently launched in April) was also on display in the exhibit hall, giving us an up-close look at the company’s disposable insulin delivery device for people with type 2 diabetes.
  • Artificial pancreas (AP) highlights included more focus on portability for outpatient studies, new products from Medtronic and Animas, and bi-hormonal control. With prototype in hand, Dr. Edward Damiano (Massachusetts General Hospital, Boston, MA) demonstrated the iPhone-based, bi-hormonal (two Tandem t:slim pumps for insulin and glucagon) artificial pancreas system that will hopefully be used in an upcoming five-day, semi- outpatient closed-loop study (222-OR). Portability for outpatient studies was also frequently mentioned at a closed-loop research meeting sponsored by the JDRF and NIDDK. The greatest minds in the field presented updates on their research, capped off by a “science fair” to show off the increasingly portable and patient-friendly systems being developed. There’s no question in our minds that the hardware aspect of AP research has really moved by leaps and bounds in the past couple years. On the more near-term front, we were very interested in two orals from Dr. Satish Garg (University of Colorado Denver, Aurora, CO) on the in-patient ASPIRE study of Medtronic’s low glucose suspend (LGS) pump/CGM system (22o- and 221-OR). (As noted above, the combination of the Enlite and MiniMed 530G pump was recently submitted to FDA; see our report at http://www.closeconcerns.com/knowledgebase/r/cbcebe6a). Most notable was the finding that “hypoglycemia begets hypoglycemia,” reminding us of the urgent need to help patients break the cycle of hypoglycemia unawareness and the potential afforded by LGS and predictive-suspend systems. The fact that Medtronic was able to submit LGS faster than expected is likely a very good sign for other pump manufacturers working on the closed loop – Animas in particular. Speaking of the latter, we finally saw data on Animas’ Hypoglycemia-Hyperglycemia Minimizer System – despite two high- carb meals and deliberate under- and over-bolusing, the system kept patients in zone (70-180 mg/dl) nearly 70% of the time, with only 0.2% of the time spent in hypoglycemia (917-P) – a very big deal. We look forward to seeing more data, especially when the device becomes more portable and is tested even more extensively. Other notable presentations focused on factors that affect closed-loop control (“controller effort” was the biggest one) and a record four AP presentations that discussed use of both insulin and glucagon. Glucagon in particular is getting a lot more attention than it used to … promising news for patients who have been working with decades-old technology but for whom new and much better “glucagon rescue” pens are in the works.
  • What is the biggest obstacle to the artificial pancreas? While the usual answer we hear at conferences is “the speed of insulin” or “CGM accuracy,” we were intrigued to hear Dr. William Tamborlane (Yale University, New Haven, CT) propose that safety actually represents the biggest obstacle (i.e., ensuring that a system’s malfunctions will not result in harmful over-delivery of insulin). We believe safety is FDA’s biggest worry as well (e.g., the delay in the Veo…), and we hope as CGM accuracy improves and algorithms get smarter, this will be less of a concern. We believe FDA may move to actually asking patients to take on more risk, in order to move things faster – it will be interesting to watch this.
  • We were encouraged to see FDA representatives more involved in device Q&A sessions than we’ve ever seen in the past. Not only were they actively asking questions and engaging with speakers, but they were publicly identifying themselves at the microphones! This is another sign to us that things may be picking up steam and moving along now, especially as JDRF and industry put more pressure on the Agency to define study requirements and expectations.

 

Table of Contents 

 

CGM and SMBG

Oral Sessions: Continuous Glucose Monitoring

IMPACT OF STUDY DESIGN AND ANALYTIC TECHNIQUES ON THE REPORTED ACCURACY OF CONTINUOUS GLUCOSE MONITORING (CGM) SYSTEMS (1-OR)

David Price, MD (Dexcom, San Diego, CA)

Dexcom’s Dr. David Price gave an excellent, Wizard-of-Oz-inspired review of how CGM data is reported, emphasizing how “turning handles and twisting levers” can skew the data. The highlight was topline data from Dexcom’s pivotal G4 study (n=72; 9,093 matched pairs): a mean absolute relative difference of 13%, 80% of points in the CEG A-Zone, and 94% of sensors lasting up to seven days. Turning to study analysis, Dr. Price advocated for taking CARE when interpreting CGM data: Calibration (frequency, method, instrument), Analytic techniques (A+B Zones vs. A Zones, median vs. mean, lag adjustment), Range (glucose and rates of change), and Excluded data (outlier values, outlier sensors, Day #1). He especially emphasized that CGM studies should present data based on a product’s intended use. We appreciated Dr. Price’s examples of how changing the data can drastically affect accuracy, especially when multiple questionable approaches are stacked (e.g., using YSI for calibration AND calibrating four times per day). Overall, this presentation was as eye opening as we expected and it was quite clear that Dexcom has a high level of confidence in the performance of their new sensor. AS a reminder, the G4 was submitted for FDA approval in 1Q12; see our most recent Dexcom earnings report at http://bit.ly/KbAhyC.

  • The pivotal study of the G4 (n=72) demonstrated a mean absolute relative difference (MARD) of 13%, with 94% of sensors lasting seven days and 80% of points in the Clarke Error Grid A-Zone. The overall percentage of sensors within 20% of reference (>80 mg/dl) and 20 mg/dl of reference (<80 mg/dl) was 82%, with 83% within 20 mg/dl at 40- 80 mg/dl. The precision between two sensors as measured by coefficient of variation was 7%. On time percentage within days (288 possible) was 97%. Dr. Price noted the improved accuracy of the G4 compared to the Seven Plus – while 73% of individual Seven Plus sensors had a mean ARD within 20%, this increased to 93% in the G4 pivotal study. The pivotal study had 9,093 matched pairs and 15% of YSI values were less than 80 mg/d.

G4 Pivotal Study Results

 

Within 20%/20 mg/dl of reference

Clarke Error Grid A-Zone

MARD %

Two daily real-time calibrations with SMBG by subjects

 

82%

 

80%

 

13

YSI Calibration

86

85

11

Four daily SMBG calibrations

 

86

 

85

 

12

Retrospective calibration (two times per day)

 

89

 

88

 

11

Retrospective YSI calibration (four times per day)

 

94

 

94

 

7

  • “An MARD of 13% does not equal a MRD of 13%.” Dr. Price characterized these common CGM accuracy acronyms as “confusing” and gave an example to illustrate how mean relative difference (MRD) can be biased. For a YSI reading of 200 mg/dl and corresponding CGM readings of 150 mg/dl and 250 mg/dl, the mean relative difference (MRD) is 0%. The MARD, however, is 25%. This illustrates the major bias of MRD, where positive and negative biases average each other out.
  • Clinical studies must be designed to reflect intended use. Dr. Price noted that studies should enroll an adequate number of intended users and patients in studies must act like patients do in real life (self-deploy sensors, self-calibrate at the labeled frequency, calibrate with glucosemoving up or down, use the sensor in values across the glucose range). Studies should also include in-clinic days throughout a sensor session (i.e., at the beginning, middle, and end of a sensor session). YSI values should also be matched to corresponding CGM values.
  • Calibration: Dr. Price showed how increasing the number of calibrations, the method of calibrations, and post processing CGM data can dramatically improve reported CGM accuracy. The first row in the table below illustrates what Dexcom observed in the G4 pivotal study, while subsequent rows illustrate the G4 pivotal study data with certain post- processing tactics.
  • Analytic: Certain analytic techniques can be misleading, such as reporting Clarke Grid A+B-Zone data vs. A-Zone data alone. Dr. Price showed how the GlucoWatch 2B and Dexcom G4 (on day seven) have similar A+B Zone data (95% and 97% respectively), while reporting the A-Zone data alone really shows the accuracy difference between the sensors: 51% of points for the GlucoWatch data vs. 85% of points for the Dexcom G4 on day seven. This difference was also highly noticeable on the Clarke Grid Plot, where the Dexcom G4 had a tight clustering around the 45-degree line, while the GlucoWatch 2B had a wider and more scattered dispersion.
  • Range: Excluding low glucose data lowers Clarke Error Grid D-Zone points and improves a sensor’s MARD. The Dexcom G4 pivotal study included points YSI values between 40 and 400 mg/dl, resulting in 9,093 matched pairs, a MARD of 13%, 80% of points in the A- Zone, and 2% of points in the D-Zone. Limiting the YSI points between 81 and 400 mg/dl reduces the number of matched pairs to 7,742, improves the MARD and A-Zone points by a single percentage point (to 12% and 81% respectively), and decreases the number of D-Zone points to just 0.5%. The reason is because Clarke Error Grid and MARD analyses amplify errors at low glucose values. For instance, at a YSI of 60 mg/dl, a CGM reading of 60 mg/dl will fall into the A- Zone, while a reading of 75 mg/dl will fall into the D-Zone. Because the denominator in the MARD calculation is the reference glucose (e.g., 60 mg/dl in the previous example), a small error in the low glucose range (e.g., 15 mg/dl), can inflate the MARD (25% in this case).
  • Excluding: Removing outlier sensors can improve a sensor’s reported MARD. Dr. Price explained that a MARD can be reported as the average error across all sensors in a study, or it could be reported as the histogram of individual sensor MARDs. Removing outlier sensor data from Seven Plus data reduced the mean ARD from 15.9% to 13% and the median ARD from 14.1% to 12.7%. He further explained that reporting medians diminishes the impact of outliers. Dr. Price argued for greater transparency in the data, especially because reporting all values better reflects the experience users will actually have.

Questions and Answers

Q: As clinicians, do you have a comparison between the G4 and Medtronic’s Enlite?

A: Medtronic’s new product performance has improved a great deal. I would say that our G4 is really revving on days four to seven. I’ve not seen data on the Enlite besides what has been presented. Based on the Enlite poster, I’m not quite sure how calibration was done, whether there was lag adjustment, etc. You’ll have to ask them at their presentation.

Q: A question on median and mean ARD. If you look at the glucose data measured by CGM, it’s not a normal distribution. You should be reporting median. You often report mean because that’s what we’re used to, but if you look at the data, it should be a median.

A: The problem with looking at the median is that it negates the impact of outlier sensors. I don’t disagree with you. But those outlier sensors are important.

Comment: So maybe we should add a measure of distribution.

 

A COMPARATIVE ANALYSIS OF THREE CONTINUOUS GLUCOSE MONITORS: NOT ALL ARE CREATED EQUAL (4-OR)

Steven J. Russell, MD (Massachusetts General Hospital, Boston, MA)

Dr. Russell presented head-to-head comparisons of multiple CGM systems (including some not yet approved in the US), with data drawn from 48-hour bi-hormonal closed-loop experiments (which offered a wide range of hyperglycemic values and rates of glucose change, but few points in the hypoglycemic range). First he showed data in which the FreeStyle Navigator performed with higher accuracy and lower variability compared to the Dexcom Seven Plus and Medtronic REAL-Time Guardian – a finding his team has discussed previously. He went on to discuss a new three-way comparison of the FreeStyle Navigator with Dexcom’s Gen 4 Sensor and Medtronic’s Enlite based on seven 48-hour experiments. The Enlite (MARD 17.2±9.5%) showed an improvement over the REAL-Time Guardian, with much of the remaining inaccuracy and variability due to some outlier values (Medtronic engineers are working on an improved algorithm, which Dr. Russell said that he and his colleagues will be testing). Meanwhile the Gen 4 (MARD 11.3±4.3%) posted better results even than the Navigator (MARD 11.9±4.3%) and – unlike the Navigator or Enlite – got better on the second day of wear, implying that subsequent days of wear might make the Gen 4 look even better (a prospect Dr. Russell’s team will “soon” investigate in five-day closed-loop studies).

  • Dr. Russell began by discussing a three-way comparison of Abbott’s FreeStyle Navigator, Dexcom’s Seven Plus, and Medtronic’s REAL-Time Guardian (12 48-hour experiments; 2,360 reference blood glucose measurements [GlucoScout]). The FreeStyle Navigator had the lowest mean absolute relative difference (MARD) and the smallest standard deviation in MARD (11.8%±3.8%), as well as the highest percentage of matched pairs in the Clarke Error Grid A Zone (81%). The results were less favorable for the Seven Plus (MARD 16.5±6.7%; CEG A 76%) and REAL-Time Guardian (MARD 20.3±6.8%; CEG A 64%). Dr. Russell also presented data on sensor reliability, which was notably higher for the Navigator and Guardian (99.8% and 98.5% of measurements captured, respectively) than for the Seven Plus (75.9%).
  • Dr. Russell further illuminated sensor performance in various less-conventional ways, including a breakdown of MARD by glucose range. (The hypoglycemic range was excluded due to the paucity of data; the closed-loop system maintained control so effectively that only 0.7% of the total measurements were below 70 mg/dl.) In each of 70-120 mg/dl, 120-180 mg/dl, and 180-250 mg/dl, the Navigator performed better than either the Seven Plus or Guardian, whereas at >250 mg/dl, the Seven Plus had the lowest MARD (the Navigator consistently under-reads in the high glucose range, Dr. Russell said). The Seven Plus had better MARD than the Guardian in all glucose ranges. Dr. Russell also presented Bland-Altman plots showing that standard deviation for the Navigator was tighter than for the other CGMs across the glycemic spectrum studied.
  • The Boston artificial pancreas researchers have conducted 17 experiments comparing just the FreeStyle Navigator (MARD 13.2±3.6%) and Medtronic’s Enlite sensor (MARD 16.0±7.4%). Dr. Russell noted that the Enlite’s standard deviation was raised by several sensors with atypically high MARD – a problem that Medtronic engineers think they can address with an improved algorithm (which Dr. Russell and his colleagues will help test once it has been developed).
  • In Dr. Russell’s group’s studies, the Dexcom Gen Four had a MARD below 10.0% on the second day of sensor wear – an improvement from day one, whereas the Navigator and Enlite performed worse on the second day. (Dr. Russell noted that his group found better Gen Four MARD than had been presented by Dexcom’s Dr. David Price earlier in the session – he attributed this in part to calibration with the GlucoScout rather than with the less-accurate BGM devices used in the Dexcom clinical study). Dr. Russell speculated that the Gen Four’s performance might be still more superior on subsequent days of wear. He and his colleagues are planning to conduct five-day closed loop experiments “soon” – we are curious whether the system will be ‘driven’ by the Navigator (the group’s historical sensor of choice) or, in light of these new data, the Gen Four.

Questions and Answers

Q: Were you able to compare during/after exercise and overnight?

A: We haven’t looked at those subsets of the data.

Q: But you have not noticed any gross differences?

A: Overnight we have tighter glucose control and lower rates of change, so lag becomes less of an issue; I would expect control to be better then. During exercise we see more rapid changes, so I would anticipate worse accuracy then.

Q: Were there any artifacts in sleep from patients rolling on the sensors?

A: In our closed-loop system, the sensors are all worn on the abdomen, so people can’t sleep on their stomach.

Dr. Roman Hovorka (University of Cambridge, Cambridge, UK): I look forward to seeing the Gen 4 in our studies.

 

ACCURACY AND LARGE INACCURACY OF TWO CONTINUOUS GLUCOSE MONITORING (CGM) SYSTEMS (3-OR)

Lalantha Leelarathna, MBBS, MRCP (University of Cambridge, UK)

Dr. Leelarathna presented Abbott FreeStyle Navigator and Dexcom Seven Plus CGM data from five closed-loop studies in a total of 52 patients at Cambridge. The accuracy and inaccuracy of both CGMs were compared on a wide variety of metrics, including MARD (9.9% for the FreeStyle Navigator vs. 12.6% for the Dexcom Seven Plus), Clarke error Grid (78% in the A-Zone for the FreeStyle Navigator vs. 71% for the Dexcom Seven Plus), frequency of large sensor errors, and error duration. Large sensor over-reading occurred two to three times more frequently with the Dexcom Seven Plus than with the FreeStyle Navigator. Additionally, at higher error levels (50% and 60%), errors of one hour or longer were absent with the FreeStyle Navigator. This led Dr. Leelarathna to conclude that the FreeStyle Navigator “may be more valuable for closed loop operation than Dexcom’s Seven Plus.” Of course, Dr. Steven Russell’s subsequent presentation demonstrated that Dexcom’s new G4 sensor seems to have closed the accuracy gap with the FreeStyle Navigator. Nevertheless, we appreciated the researchers and session’s focus on more than just top-line CGM accuracy data – reducing large sensor errors will not only improve AP performance, but we suspect it might also lead less patients to become frustrated and quit using CGM. We were particularly intrigued by the multiple questions from FDA during the Q&A. It seems the agency is wrestling with this idea of statistically defining and analyzing large sensor error data.

Questions and Answers

Q: Have you modeled the number of severe lows that would have occurred as a result of over readings?

A: We do have some simulation data. After 40% or more errors, that would cause hypoglycemia.

Q: Did you adjust for the fact that Navigator doesn’t show the very first few hours of data?

A: We used the one-hour Navigator in our studies. We also inserted both sensors one day before. They were identically 24 hours into their sensor life.

Q: How did you calculate error duration? What happens if a sensor went from 40% to 50% back to 40% error? Did you take individual time points and call that a minute or was it continuous duration? Did you use arterialized or venous blood?

A: We inter-collated into one-minute data. We took YSI every hour, but inter-collated into one-minute data.

Q: What if it moved back and forth?

A: That would be classified as a second error. If an error came out of an error zone, the error ended. The second time we went back it was counted as a second error. And we used venous plasma glucose.

Q: How did you define an event? Say you went from 30% to 40% to 30% to 40% error. Would three separate events be counted?

A: That’s correct.

Q: And the event was a CGM point compared to blood glucose?

A: Yes, reference glucose vs. CGM glucose.

TRANSDERMAL CONTINUOUS GLUCOSE MONITORING FOLLOWING SKIN PERMEATION (7-OR)

Wayne Menzie (Echo Therapeutics, Philadelphia, PA)

Wayne Menzie discussed the performance of Echo’s transcutaneous continuous glucose monitoring device, the Symphony tCGM, in a 24-hour pilot study of people with diabetes (n=20; 12 with type 1 diabetes, 8 male, mean age 51.5); sensors were calibrated four times a day. As announced in December 2011, mean absolute relative difference was 12.6%, and 94.4% of matched data pairs were within the A zone of the Continuous Glucose Error Grid Analysis. Mr. Menzie noted that the sensors used in the study had a bimodal distribution of accuracy: the 15 best-performing sensors had a MARD of 11.3%, whereas the five worst-performing sensors had MARD of 20.1% (Echo’s team has identified the source of inaccuracy in these sensors as a hardware problem and believe they have addressed the issue). Next steps include pilot studies in the ICU (ongoing), tests with additional skin types and body locations (the study in diabetes included arms and abdomens), more data in the hypoglycemic range (Mr. Menzie acknowledged that this was still a “blind spot” in Echo’s dataset), studies that last beyond 24 hours, and further studies of potential interferents (though Mr. Menzie said that neither acetaminophen nor ascorbic acid seem to significantly interfere with the Symphony, based on preliminary research).

  • Although Echo’s near-term target market is critical care patients, healthy people with diabetes offered the company a chance to test the system across a wide range of glucose levels (2,680 measurements with mean glucose 157±62 mg/dl; range 42-333 mg/dl).

Questions and Answers

Q: Do you have data on repeated use?

A: Some subjects have been volunteers in multiple studies, but these are usually separated by months. The site itself takes several days to regenerate. Skin permeation is a precise technique, so we can’t go to same site until three-to-four days have passed.

Q: Have you seen any adverse events?

A: No, the worst thing we’ve seen is minor skin irritation; the biggest cause of irritation is actually the adhesive.

Q: On the last slide you presented data that you said was without calibration… how do you account for background current and interference?

A: Calibration was done every four hours – since blood glucose is taken more frequently in hospitals, we would look to those measurements to improve performance. Our calibration algorithm is still under review, however. No specific measures taken other than design of device, there was no signal processing or correction for interference. In limited studies we haven’t seen that problem but we will study the interference question further; we know this is issue especially for the hospital.

Q: I recommend you study ascorbic acid and acetaminophen to see what happens with those.

A: We’ve tried both of those preliminarily and haven’t seen problems.

 

A1C AND MEAN GLUCOSE (MG) IN INSULIN TREATED DIABETES USING THE DEXCOM SEVENPLUS CONTINUOUS GLUCOSE MONITOR (CGM): CORRELATION AND INTRA- PATIENT CONSISTENCY OVER TIME (2-OR)

Nicholas Argento, MD (Maryland Endocrine and Diabetes Center, Columbia, MD)

Dr. Argento discussed “high glycolators” and “low glycolators” – people whose CGM-measured mean glucose (CMG) differs widely from the estimated average glucose (eAG) based on A1c. To study the relationship between CMG and A1c, his team analyzed A1c and CGM (Dexcom Seven Plus) data from several dozen patients in Dr. Argento’s own practice. Similar to the ADAG and JDRF CGM studies, on average a CMG of 154 mg/dl translated to an A1c of 7.0% – a CMG/A1c ratio of 21.7. But (also as seen in the JDRF CGM trial), the CMG/A1c ratio varied widely across the population and was stable within individuals. This phenomenon means that “high glycolators” (patients in the highest decile of CMG/A1c ratio – i.e., below 19.2 mg/dl%) actually have lower mean glucose than their A1c would indicate, and so are in danger of frequent hypoglycemia. By contrast, “low glycolators” (patients in the lowest decile of CMG/A1c ratio – i.e., above 24.9 mg/dl%) actually have higher mean glucose than their A1c would indicate, and so are at much greater risk of complications than they might realize. Dr. Argento proposed that in both high and low glycolators, CGM-measured mean glucose is a better measure than A1c for setting glycemic targets.

Questions and Answers

Q: Could you comment on mean glucose vs. A1c as a target? We know that at the same glucose level, A1cs are higher in older people and in blacks.

A: Black patients in our study appeared to have residual negative – A1c is higher than would be expected from mean glucose. I’m not saying we shouldn’t use a1c; as best we can tell it is a surrogate marker of mean glucose, but it seems not to be a great marker in all patients. Ten genetic loci with high variability have been found to influence A1c; only three of those seem to have to do with glycemic control. One would think that other factors that influence A1c, such as iron metabolism and red cell life, don’t have effect on diabetic complications. Should we look only at mean glucose? No, because variability is important too – of course, CGM is also probably the best way to look at this too. If you are outlier in CCMG/A1c, you may want to modify the way your glycemic targets are set.

Q: I am trying to understand the significance of glycation. Maybe if you are a low glycator, you would glycate renal proteins less also.

A: That is a great question. If you make hemoglobin at a lower rate, perhaps you are protected elsewhere. But there so many other factors – like iron metabolism and red cell life – seem that they would be unlikely to protect kidneys or endothelial tissue. There has been talk of this, but I am not aware of evidence. Perhaps some patients have partial protection in some tissues; perhaps others have none. I think the way to go is mean glucose, since we know what that means.

 

LONG-TERM EFFECTS OF SENSOR-AUGMENTED PUMP THERAPY IN TYPE 1 DIABETES: A 3-YEAR FOLLOW-UP STUDY (8-OR)

Signe Schmidt, MD (Copenhagen University Hospital, Hvidovre, Denmark)

Dr. Signe Schmidt presented three-year follow-up data on 24 Danish patients from the Eurhythmics trial of sensor-augmented pumping (SAP). The patients that were still using SAP at three years (n=16) maintained the vast majority of the glycemic improvement that they had achieved during the 26-week study (three-year A1c below 7.5%, from a baseline of roughly 8.5%). These patients also reported improvements in the Diabetes Treatment Satisfaction Questionnaire (p<0.01), the Problem Areas in Diabetes survey (p<0.02), and the Hypoglycemia Fear Survey (not statistically significant). Notably, however, the patients that stopped using CGM and switched to either pumping alone (n=4) or MDI (n=2) achieved extremely similar three-year glycemic control. Dr. Schmidt noted that the small sample size makes true analysis difficult, but she speculated that perhaps these six patients had gained important diabetes insights from their time wearing sensors and that they continued to apply these insights even after ceasing CGM use (a phenomenon that Dr. Robert Vigersky, who commented during Q&A, reported in a study of intermittent CGM use in patients with type 2 diabetes not using mealtime insulin [Vigersky et al., Diabetes Care 2011]).

Questions and Answers

Q: There was a striking lack of improvement in the hypoglycemic fear index, even though we often think of SAP as a way to address hypoglycemia. Why do you think that was the one measure that didn’t improve?

A: Actually it did improve, the improvement was just non-significant. This depends on sample size.

Q: In the SWITCH study, people using SAP who stopped wearing the sensor saw a decline in their A1c.

A: SWITCH was designed to address the question of SAP vs. pumping; it suggests that there is a difference between them.

Q: I have two questions. In the initial randomization between MDI and SAP, did the MDI patients get as much contact and education as the SAP group?

A: I was not part of the Eurhythmics team, so I am not sure how randomization worked. I can say that when MDI patients started SAP in our clinic, they got the same education as the SAP group in the Eurhythmics trial.

(Eurhythmics Investigator): We used much more time to train the SAP patients compared to MDI – it was obvious that we needed to train them.

Q: When you switch from MDI to SAP you make two changes. Can you speculate which part of SAP was more important?

A: To that question I can only refer to meta-analysis by John Pickup, which suggested that both components have great impact.

 

GLYCEMIC VARIABILITY IS HIGHER IN TYPE 1 DIABETIC PATIENTS WITH MICROVASCULAR COMPLICATIONS IRRESPECTIVE OF GLYCEMIC CONTROL (5-OR)

Jan Soupal, MD (Charles University, Prague, Czech Republic)

Dr. Soupal detailed an interesting CGM study comparing glycemic variability in type 1 patients with and without microvascular (MVC) complications. Thirty-two patients (mean age: 43 years, mean A1c: 9.5%, mean duration of diabetes: 19 years, n=16 with MVC) wore blinded CGM for two weeks and performed at least four fingersticks per day. There were no significant differences in baseline criteria between patients with and without MVCs. Glycemic variability (SD, MAGE) calculated from CGM was significantly higher in type 1 diabetes patients with retinopathy (p=0.02), microalbuminuria (p=0.035), and impaired vibration perception threshold (p=0.01). Moreover, patients with any MVC had significantly higher glycemic variability than patients without complications, although they didn’t differ in glycemic control (p=0.019). Of course, this data is merely correlational and does not demonstrate that higher glycemic variability caused the complications. Nevertheless, we do think it is interesting and we wish that a real trial could be done long-term that would show the impact of glycemic variability on long-term outcomes. In our view, most notable was that when glycemic variability was calculated using SMBG data from the study, these relationships did not hold. At past conferences, we’ve often seen SMBG data used to calculate glycemic variability statistics, though we’ve heard a few speakers characterize this as inappropriate; Dr. Soupal’s analysis may indeed suggest it is questionable, since patients may be paying more attention to their blood glucose when they know they will be tested. Overall, we’re glad to see increasing focus on these issues and we look forward to prospective glycemic variability studies like Dr. Irl Hirsch’s FLAT-SUGAR, which could turn out to be a landmark study in some years from now.

Questions and Answers

Q: Complications take years to develop. Many become born again diabetics when complications show up. Did you look at historical values?

A: That’s an important limitation of each observational, case control study. We don’t know the past. It was a limitation of this study and we have to think about it. This study should be considered for a larger study. It should be multicenter and obtain a lot of patients. It must be prospective. That’s the only way to get definitive results – does glycemic variability really increase the risk of microvascular complications. We can demonstrate correlation, but not causation.

Q: Did you do subgroup analyses? Was there a significant correlation with high A1c patients?

A: We did multivariate analyses but it didn’t affect the results. The group of patients was very homogenous.

Q: Did you adjust the models for A1c?

A: Yes, we did multivariate analysis.

 

PERFORMANCE OF A MICRODIALYSIS-BASED CONTINUOUS GLUCOSE MONITORING (CGM) SYSTEM (6-OR)

Eric Zijlstra, PhD (Profil Institute for Metabolic Science, Neuss, Germany)

Dr. Zijlstra presented the results from a 10-48-hour study of an intravenous microdialysis-based CGM in 21 healthy individuals. The system achieved an MARD of 9.4% and 91.4% of points in the Clarke Error Grid Zone A. Calibration occurred once per day and the glucose range was 42-267 mg/dl. An advantage of the system is that there is no blood loss to the patient. The accuracy looks decent, but nowhere approaching the Dexcom/Edwards GlucoClear2 data we saw at ATTD (MARD of 5.2%).

  • Dr. Zijlstra described the intravenous microdialysis-based CGM used in the study, which measures glucose every one to two minutes without blood loss. To use the microdialysis unit, a standard blood catheter is inserted into the vein of the forearm. For each measurement, the system automatically draws blood into the sampling line. Glucose from the blood diffuses over a thin membrane, is perfused with saline solution, and transported to a glucose sensor for monitoring. The blood is then flushed back and returned to the patient. A glucose value is recorded every one to two minutes.
  • The accuracy and reliability of the CGM was assessed in 21 healthy volunteers (mean age: 29 years, mean BMI: 23.7 kg/m2). Experiments were generally ten hours long and four volunteers did 48 hours. The glucose sensor was calibrated once every 24 hours. The system was calibrated before the experiment using a two-point calibration. Reference blood samples were taken manually and analyzed using a laboratory glucose analyzer every 10-60 minutes. The volunteers consumed meals or glucose was administered orally or intravenously to analyze the accuracy of the CGM system over a range of blood glucose concentrations.
  • The system achieved a mean absolute relative deviation of 9.4% (n=1796 matched pairs; glucose range of 42-267 mg/dl). Mean absolute deviation was 10.7 mg/dl. At glucose values<75 mg/dl, 94.6% of points were within 15 mg/dl. For glucose values >75 mg/dl, 90.4% of points were within 20% of reference. A Clarke Error Grid Analysis showed 91.4% of points in Zone A, 8% in Zone B, 0.6% in Zone D, and 0.1% in Zone E.

Reference BG

N

Mean Absolute Deviation

Mean Absolute Relative Deviation

All (42-267 mg/dl)

1796

10.7 mg/dl

9.4%

<70 mg/dl

191

6.7 mg/dl

11.0%

70-180 mg/dl

1519

9.6 mg/dl

8.8%

>180 mg/dl

154

25.9 mg/dl

12.6%

48 hours (53-184 mg/dl)

 

367

 

8.2 mg/dl

 

7.8%

Questions and Answers

Dr. Bruce Buckingham (Stanford University, Stanford, CA): Is there a lag with this system?

A: There’s a little lag time due to the transportation of the sample to the sensor. Probably two minutes lag time.

Dr. Steven Russell (Massachusetts General Hospital, Boston, MA): How did you avoid blood clotting?

A: In this setup, the microdialysis is integrated into the sampling line. We had two-way pumps. The pump would draw blood for 30 seconds and then flush. The microdialysis membrane would be flushed after every measurement.

Q: These were healthy subjects. Did you induce hypoglycemia?

A: Yes, we did induce hypoglycemia.

Q: How much dialysate was done every day? Is this a glucose oxidase sensor?

A: Yes, it’s glucose oxidase. On the amount of dialysate, I’m not sure. It’s relatively little. I think on average 100-125 microliters per minute.

 

Oral Sessions: Expanding the Domains of Diabetes Education

IMPACT OF A POCKET INSULIN DOSING GUIDE ON UTILIZATION OF BASAL/BOLUS INSULIN BY INTERNAL MEDICINE RESIDENT PHYSICIANS (77-OR)

Michael Jakoby, MD, MBA (Southern Illinois University School of Medicine, Springfield, IL)

Dr. Jakoby discussed Southern Illinois University’s efforts to distribute a pocket insulin-dosing guide among internal residents, to increase the appropriate use of basal/bolus insulin therapy (as opposed to sliding scale insulin therapy, which Dr. Jakoby found was used in 90% of patients in the Carle healthcare system in Urbana, IL – scary!). A pilot program including all internal medicine residents began in November 2010, and an eight-month extension study began in July 2011 (when new internal medicine interns arrived and a second staff presentation on the guides was given). Compared to historical controls (November 2009 to October 2010), distribution of the pocket guides significantly increased basal-bolus insulin prescriptions by 2.5-fold on average. This change appeared durable, though the highest rates of prescription seemed to occur after the staff presentations in November 2010 and July 2011. No increase in hypoglycemia was observed, and glycemic control among patients that had spent several days in the hospital was better following the introduction of the guide. Also following the guide, statistically non-significant improvements were seen in length of stay (4.8 vs. 5.7 days, p=0.08). Next steps include refining the card’s orders (no modifications were made during the study itself, to preserve the integrity of the data), increasing acceptance of the system among physicians and other clinicians at SIU and beyond, and developing a computer-based algorithm based on the card (Dr. Jakoby noted that pocket-based guides are not as useful in this modern age, when physicians input orders via computer and thus rarely reach into their pockets for pens).

Questions and Answers

Q (Physician from the UK): Do I understand that for patients that come in on oral hypoglycemics, you would recommend they go on basal-bolus insulin in the hospital?

A: Unequivocally. We have different philosophies on different sides of the Atlantic. Dr. Umpierrez at Emory and Dr. Baldwin at Rush have shown that patients are better managed on insulin than on orals.

Q: Do you think your junior doctors are safe enough?

A: When we give them tools, they are certainly not as successful as the other authors I’ve cited, but there was steady improvement over time without an increase in hypoglycemia. I would posit that our results show house staff can be trained to manage diabetes with reasonable efficacy.

Q: We did a snapshot day where we looked at all people in hospital; glycemic control for those on insulin was a very big problem. This is across the whole of England, with 12,000 patients. I have concerns about our current use of insulin regimes in the UK; that is all I will say. I hope you are doing better.

A: We have a database of about 1,200 patients and data from my time at the University of Illinois and Champaign-Urbana, and we are teasing this issue out. With regard to patients that remain on sulfonylureas, they have actually higher rates of hypoglycemia than those on basal-bolus insulin.

Q: I appreciate your study of one area where failure can occur, residents. What about the nursing staff? Sometimes the basal insulin will be incorrectly held because glucose is normal. Have you factored that in?

A: We did not. We will go back and look at that. Obviously successful inpatient management depends on many factors. We have a team with an NP, educator, dietitian; we are tracking that and finding better compliance. Another problem we identified is poor coordination between delivery of trays and prandial insulin dosing. Tray deliverers often drop off the tray and then run to the next patient without notifying nurses that the tray has arrived. We will be trying a new system to link the processes better.

Q: What regimes are patients going home on? Did you look at all at readmission at patients that had used this kind of dosing?

A: We are fairly aggressive about sending patients home on basal-bolus insulin. We sent about 85% home on basal-bolus insulin and tracked them with the Carle Clinic system. We found compliance was very high and A1c improved from 8.1% to 7.2% over three months. (Editor’s note – wow!) As mentioned earlier in the session, there is nothing like a severe acute illness to clarify someone’s thinking, and a lot of downtime during the day in the hospital can be used for education.

Q: Is this analysis published anywhere?

A: That is next on the agenda.

Q: We tried a similar pocket dosing guide but found we had to keep updating the card – for instance, to include specific guidance on renal failure.

A: We had instructions for how to handle patients with renal failure on the order set but not card; we will change this in version 2.0, but for the integrity of the study we kept the card the same throughout the study duration. Implementing the dosage guide more widely will be challenging – we have to convince a variety of groups that it is in their self-interest. So first we are trying to demonstrate success at SIU.

 

EDUCATOR USE OF MASKED CONTINUOUS GLUCOSE MONITORING DEVICE (CGM) IN A CLINIC POPULATION OF YOUTH WITH TYPE 1 DIABETES (T1D) (75-OR)

Kerry M. Milaszewski, BS, RN, CDE (Joslin Diabetes Center, Boston, MA)

Kerry Milaszewski analyzed the A1c-lowering effects of a blinded CGM intervention among youth with type 1 diabetes (n=122; mean age 14 years, A1c 8.5%, 61% pumpers, diabetes duration 7.5 years). During their three-days of blinded CGM use, patients and their families also maintained a log of glucose values, insulin, food, and activity. Based on the log and the sensor data, Joslin clinicians gave patients therapeutic recommendations (mean 3.1 recommendations per patient). Although no significant A1c effect was seen in the population as a whole two-to-three months later, 39 patients improved their A1c by at least 0.5%. Compared to those with lesser benefits, these ‘responders’ were older (15.5 vs. 13.9 years), with longer diabetes duration (8.7 vs. 6.9 years) and higher baseline A1c (8.9% vs. 8.2%). Improvement was especially likely among those that received the recommendation to use advanced bolus techniques and/or to attend to active insulin – these people were fourfold likelier to improve significantly. Notably, advanced bolus techniques were taught to both pumps and MDI users – mode of insulin delivery did not correlate with success in the Joslin intervention. We would have been interested to see follow-up CGM data on other glycemic measures besides A1c (e.g., hypoglycemia, hyperglycemia, glycemic variability) – as suggested during Q&A, the three-day wear intervention may well have improved patients’ time in zone even though it did not influence mean A1c.

Questions and Answers

Q: Did standard deviation get narrower? Did they have less hypo reactions?

Ms. Milaszewski: We did not look at hypoglycemia at all, and I’m not sure of answer in terms of SD.

 

Oral Sessions: Can We Rein in the Costs of Diabetes with Better Diabetes Care?

SELF-MONITORING BLOOD GLUCOSE TEST STRIP UTILIZATION IN CANADA (132-OR)

Jason Yeaw, MD (IMS Consulting Group, Redwood City, North Carolina)

Dr. Yeaw presented his study of blood glucose test strip utilization in Canadian diabetes patients with insulin injections. The study found that the average Canadian with diabetes on insulin injections uses 1,094 test strips per year (~three per day), which cost $860 Canadian dollars. Test strips on average make up 41.6% of total diabetes-related pharmacy costs.

  • The Canadian patients in this study averaged three test strips per day. (We note that this meets the ADA’s recommendation that MDI users test three or more times daily.) The average cost per strip was $0.79 CAD. As described by Dr. Yeaw, the study used IMS Brogan drug plan data for the period of July 1, 2006 to June 30, 2010, and only considered patients who were expected to self-monitor blood glucose. The study included 142,551 patients with type 1 or 2 diabetes who had at least two prescriptions of insulins (basal only, basal-bolus, bolus only, or premix).
  Number of Subjects Number of test strips per patient per year Mean annual test strip cost per patient (Canadian dollars) Test strips as a proportion of total diabetes related pharmacy cost
Overall 142,551 1,094 $860 41.6%
Basal 45,003 935 $740 33.7%
Basal-bolus 38,553 1,324 $1,056   41.5%
Bolus only  43,470  1,413 $1,112   52.8%
Premix  15,525  890  $678  41.4%

 

​Questions and Answers

Q: Can you tell us anything about how the strips are actually used in the patients who are only using basal insulin? Have you had focus groups, conversations with providers, etc., to look at what strips are used for?

A: I agree this is a helpful extension to this study and a good future project.

Q: In Europe, it’s about one Euro also per strip. In Canada you pay almost one dollar. How can we push for better or lower prices?

A: We really don’t have an idea why the price is what it is. The technology is indeed no longer new or complex. The price does seem fairly uniform across the world, though.

Q: Is the share of costs similar in the US? Is there any effort to rein that in? 30-40% seems very high.

A: Yes, I remember that a study in the US showed that the costs of strips there is also in the 30-40% range Of course, the percentage is of diabetes pharmacy costs, which are basically test strips and insulins. I’m not aware of any current endeavors to rein in the costs of blood glucose test strips.

Posters

ACCURACY AND ACCEPTABILITY OF THE 6-DAY ENLITE CONTINUOUS SUBCUTANEOUS GLUCOSE SENSOR (30-LB)

Timothy Bailey, Ronald Brazg, Mark Christiansen, Andrew Ahmann, Robert Henry, Satish Garg, Elaine Watkins, Francine Kaufman

This poster reported results from the pivotal six-day trial of Medtronic’s Enlite CGM sensor, which was conducted at seven US research centers and enrolled 90 adults with type 1 (n=65) or type 2 (n=25) diabetes. For the 61 patients that wore the Enlite for the full six days, accuracy data were reported from the trial’s frequent sample testing (FST) periods – i.e., roughly 12-hour in-clinic visits on days one, three, and six when YSI reference blood glucose values were taken. Two calibration regimes were compared: “actual use calibration” (three-to-four prescribed calibrations per day, mean calibration frequency 2.8±0.9 per day) and “minimal calibration” (calibration prescribed every 12 hours, mean calibration frequency 1.2±0.9 per day). Accuracy results appeared slightly better with the actual use calibration than the minimal calibration (MARD 13.6% vs. 14.7%; consensus error grid A-zone 86% vs. 81%) and when calibration occurred during a glucose rate of change that was slow (<1 mg/dl) rather than rapid (≥2 mg/dl) (MARD 13.6% vs. 16.3%). We thought these accuracy results were favorable overall, though we are uncertain if sensors were calibrated prospectively or retrospectively (we assume the former but the poster did not specify). Mean responses to a seven-point-scale survey indicated that study participants had favorable views on the Enlite’s ease of insertion (5.9), comfort (6), and ease of use (5.8), and patients also reported high likelihood of recommending the sensor to others (5.8).

  • The pivotal trial of the 6-day Enlite continuous glucose monitoring (CGM) sensor enrolled 90 adults with diabetes (65 with type 1 and 25 with type 2 diabetes; mean age 44 years, range 18-71 years) wearing one to two sensors on their abdomen. Sixty-one of the 90 participants wore Enlite sensor(s) on their abdomens for six days, while 29 patients only wore buttock sensors. The poster only considers data from the 61 patients who wore abdominal sensors. Of this group, 29 wore two abdominal sensors simultaneously and 32 wore one abdominal sensor.
  • This poster described results from clinic visits for frequent sample testing (FST), which occurred on days one, three, and six, lasted roughly 12 hours each, and included intentionally induced hypo- and hyperglycemia. Sensors were calibrated either three-to-four times per day (“actual use calibration”) or every 12 hours (“minimal calibration”), and calibration could occur even during rapid glucose changes. (We are not sure whether calibration was conducted retrospectively or prospectively.) A Yellow Spring Instruments (YSI) analyzer was used to collect reference plasma glucose values every 15 minutes and every five minutes if blood glucose <75 mg/dl. A 0-5 minute (left inclusive and right exclusive) window was used for sensor values paired with YSI values.
  • Accuracy during FST was reported in terms of mean absolute relative difference (MARD), defined as the mean absolute difference between paired sensor glucose and blood glucose values (MAD) divided by the mean blood glucose, with the result multiplied by 100% to give a percentage. The poster also included data on bias, defined as the mean difference between paired sensor glucose and blood glucose values. The standard deviations for MARD and bias were calculated per paired point.
  • Performance data during FST were broken down by a variety of categories, as shown below. Accuracy with minimal calibration (mean MARD 14.7%, mean bias -2.4 mg/dl) was only slightly lower than with actual use calibration (mean MARD 13.6%, mean bias -1.2 mg/dl). We see this result as encouraging, since many patients’ real-world use may be more similar to the minimal calibration regime (1.2 calibrations a day, rather than 2.8 for the actual use condition). Regardless of calibration condition, MARD during FST was lower on day three than either day one or day six. As for glucose ranges, MARD was greatest during hypoglycemia and MAD was lowest, as would be expected. The standard deviation of bias was quite high (above 25 mg/dl overall, for each calibration condition), which we interpret to mean that some sensors were consistently biased high and others were consistently biased low. The high variation in bias was in line with Dr. Steven Russell’s comments during his ADA 2012 presentation about a head-to-head study of next-gen CGM sensors (4-OR). As a reminder, Dr. Russell said that several Enlite sensors in his group’s study were outliers with much higher MARD than usual, and he added that Medtronic engineers think this problem can be addressed with an improved algorithm.
  • By rate of change (ROC) between SMBG calibrations, during FST

 

Slow

Moderate

Rapid

Glucose ROC (mg/dl/min)

 

<1

 

1 to <2

 

≥2

n

5,745

1,282

381

MARD, % (mean ± SD)

13.6 ± 13.9

12.9 ± 11.8

16.3 ± 14.8

MARD, % (median)

10.1

9.6

12.9

  • By day, actual use calibration (calibration frequency 2.8±0.9/day), during FST. Consensus error grid scores for all actual-use-calibration data pairs were as follows: 86% A, 13% B, 1.1% C, 0% D.

 

Day 1

Day 3

Day 6

Overall

 

MARD, %

Bias, mg/dl

MARD, %

Bias, mg/dl

MARD, %

Bias,

mg/dl

MARD, %

Bias,

mg/dl

Mean ± SD

 

15.9 ± 14.9

 

-0.3 ± 28.3

 

11.8 ± 10.9

 

0.9 ± 22.8

 

13.2 ± 14.4

-4.3 ±

34.3

 

13.6 ± 13.6

-1.2 ±

28.9

Median

12.2

0.5

9.1

1.5

9.6

1

10.1

1

  • By day, minimal calibration (calibration frequency 1.2±0.9/day), during FST. Consensus error grid scores for all minimal-calibration data pairs were as follows: 81% A, 17% B, 1.3% C, 0.1% D.

 

Day 1

Day 3

Day 6

Overall

 

MARD,

%

Bias, mg/dl

MARD, %

Bias, mg/dl

MARD, %

Bias,

mg/dl

MARD, %

Bias,

mg/dl

Mean ± SD

15.3 ±

14.0

 

0.3 ± 27.8

 

13.4 ± 11.6

 

-0.2 ± 27.1

 

15.5 ± 15.4

 

-7.8 ± 43.1

 

14.7 ± 13.7

-2.4 ±

33.3

Median

11.6

-0.3

10.6

2.5

10.4

1

10.8

1

  • By glucose range, actual use calibration, during FST. The percentage of sensor readings during the in-clinic portion broke down as follows: 9.6% (<75 mg/dl), 53.3% (75-180 mg/dl), and 37.1% (>180 mg/dl).

YSI

Reference Range

 

 

MARD, %

 

MAD, mg/dl

 

Bias, mg/dl

 

≤75 mg/dl

Mean ± SD

17.4 ± 17.9

10.8 ± 10.9

5.6 ± 14.3

Median

13.5

8.5

4.4

>75-180

mg/dl

Mean ± SD

12.6 ± 12.0

15.3 ± 14.9

2.3 ± 21.2

Median

9.4

11.3

2

 

>180 mg/dl

Mean ± SD

12.0 ± 11.0

31.0 ± 31.6

-12.6 ± 42.5

Median

9

22

-9

 

Overall

Mean ± SD

13.6 ± 13.6

18.7 ± 22.0

-1.2 ± 28.9

Median

10.1

12

1

  • By glucose range, minimal calibration, during FST

YSI

Reference Range

 

 

MARD, %

 

MAD, mg/dl

 

Bias, mg/dl

 

≤75 mg/dl

Mean ± SD

18.4 ± 15.8

11.5 ± 9.6

3.2 ± 14.6

Median

14.9

9.4

1.8

>75-180

mg/dl

Mean ± SD

14.2 ± 12.4

17.3 ± 15.1

2.2 ± 22.8

Median

10.7

13.2

3.5

 

>180 mg/dl

Mean ± SD

12.8 ± 13.4

32.8 ± 38.4

-12.7 ± 48.9

Median

8.4

21

-5

 

Overall

Mean ± SD

14.7 ± 13.7

21.0 ± 26.0

-2.4 ± 33.3

Median

10.8

13.5

1

  • Mean responses to a 30-question survey indicated patient satisfaction across all  four of the domains assessed (on a scale of 1 to 7): ease of insertion (5.9), comfort (6), ease of use (5.9), and likelihood of recommending Enlite to others (5.8). The poster included a table of representative statements and responses ranging from 1 (“strongly disagree”) to 7 (“strongly agree”). In the absence of comparison to Medtronic’s Sof-sensor and/or another sensor, these results are somewhat difficult to interpret, but attitudes certainly seem to be favorable overall.

Statement

n

Mean

SD

Median

The sensor was comfortable under my skin.

89

6.0

1.4

6

I did not feel the sensor underneath my skin

89

5.8

1.7

7

The sensor was easy to remove from my skin

89

6.3

1.2

7

The sensor started up reliably

89

6.1

1.3

7

The sensor performed well on the final day

88

5.7

1.8

6

The sensor insertion device was easy to use

90

5.7

1.5

6

Sensor insertion was pain-free

90

5.5

1.7

6

I like that I do not have to see the needle

90

4.7

1.8

4

I like that the retractable needle protects me from injury

90

5.9

1.2

6

Meet the Expert Sessions

CONTINUOUS GLUCOSE MONITORING CHALLENGES

Howard Wolpert, MD (Joslin Diabetes Center, Boston, MA)

Dr. Wolpert gave a whirlwind tour of interpreting CGM downloads and insulin delivery data, including an excellent handout illustrating his tips, thought process, and what he sees most commonly in his patients (those interested in a copy can email him). Dr. Wolpert had a number of recommendations for clinicians to “make sense” of CGM reports, including: (1) identifying frequent hyperglycemia/hypoglycemia excursions and occurrences of variability using the hourly glucose statistics report; (2) checking for rebound hyperglycemia due to reduced basal rates or overreaction to CGM lag time; (3) evaluating whether boluses are sufficient to correct hyperglycemia; (4) checking to see if delayed infusion set changes are contributing to high or erratic glucose; and (5) identifying early/late postprandial hyperglycemia due to meal content and bolus timing. For optimizing alarm thresholds, Dr. Wolpert recommended a two-step process: deciding on initial settings (usually a conservative 55-60 mg/dl for lows and 250 mg/dl or more for highs) and refining them over time to optimize benefits. We were struck by the number of examples that required both CGM and insulin delivery data – in addition to Medtronic’s pumps and well-regarded CareLink software, we look forward to more integrated systems coming to market from Insulet/Dexcom, Animas/Dexcom, Tandem/Dexcom, and Roche/Dexcom.

  • Hourly glucose statistics reports can “give important insights into highs and lows.” Dr. Wolpert showed the Dexcom Hourly Statistics report, which summarizes two or more weeks of data into 24 hourly bars with mean, median, and 25-75th percentile glucose values. Dr. Wolpert first recommends drawing in target thresholds (e.g., 80 mg/dl and 180 mg/dl). Then, one can clearly see where hourly bars exceed the threshold lines, indicating frequent occurrences of hypoglycemia and hyperglycemia. The report is also useful for identifying times of the day where there is high variability, prompting a discussion with the patient on contributing factors. Finally, if an HCP identifies frequent hyperglycemia, then he or she can easily look at the following period to assess for subsequent hypoglycemia. If there is none, the insulin dose can be safely increased.
  • CGM data can help identify some of main reasons for exaggerated rebounds from hypoglycemia: patient overreaction to CGM lag time and temp basals/pump suspensions. Dr. Wolpert showed the Medtronic Quick-View summary, which combines fingerstick, CGM, and insulin delivery data. In one example, hypoglycemia consistently prompted a patient to reduce his basal rates, resulting in a marked hyperglycemic rebound. Dr. Wolpert characterized this behavior as “a common practice in CGM users that leads to glycemic variability.” Turning to lag time, he explained that many patients treat a low with carbs, see no upward trend on the CGM, and eat further carbs. The problem is compounded by neurocognitive function, where patients continue to “feel” low (and the CGM continues to read low) even though their glucose has normalized. To prevent this type of rebound hyperglycemia, Dr. Wolpert recommends patients take a fingerstick before eating more carbs.
  • Looking at a patient’s basal/bolus proportion and the frequency of infusion set changes is also helpful to drill down into causes of hyperglycemia. Dr. Wolpert again used the Medtronic Quick-View summary to illustrate how insulin delivery data and CGM data can be mutually informative. He showed what to look for when patients under-bolus despite persistent hyperglycemia. Reasons for this behavior might include fear of hypoglycemia and concerns about weight gain. Dr. Wolpert specifically mentioned a study from Drs. Bill Polonsky and Barbara Anderson conducted in Joslin waiting rooms. One-third of women surveyed acknowledged intentionally under-dosing insulin due to concerns about weight gain. Turning to infusion set changes, Dr. Wolpert explained that delayed replacement of infusion sets is sometimes a “common cause of high/erratic glucoses and is a teachable moment!” If this is the case in certain patients, Dr. Wolpert recommends changing sets more regularly.
  • Key steps in the initial CGM/pump download review:
    • 1. Check priming history to assess frequency of infusion system change
    • 2. Check bolus history to detect possible missed meal boluses (“surprisingly common”)
    • 3. Check percentage of basal to bolus insulin.
      • If frequent hypoglycemia with basal > bolus, it may indicate that bolus doses are frequently being missed.
      • If frequent hyperglycemia with basal < bolus, it may indicate that basals are too low.
      • If frequent hypoglycemia with basal > bolus, it may indicate that high basals are contributing to hypoglycemia.
      • If frequent hypoglycemia with basal < bolus, it may indicate that excessive boluses are contributing to hypos.
    • 4. Check for pump suspension or basal rate reduction.
  • “When it comes to optimizing post-breakfast control, it comes down to looking at the food.” CGM can offer excellent insights into early and late postprandial hyperglycemia – common causes include meal type (carb and fat content) and bolus timing. Dr. Wolpert honed in on the breakfast period, where patients are more insulin resistant and typically eat high carb content, high glycemic index meals (“a carb isn’t a carb”). He recommends switching to lower glycemic index carbs, increasing protein/fat content of the meal, bolusing before breakfast (often hard for people to do in practice), or taking a larger bolus and cutting late-morning basal rates (what we’ve sometimes heard characterized as a “super bolus”). The latter allows the tail end of the bolus to cover the late-morning basal rate, preventing delayed hypoglycemia.
  • “It’s more than just fat delaying gastric emptying. There is a different insulin to carb ratio for a high fat meal vs. low fat meal.” Dr. Wolpert reviewed the importance of dietary fat on postprandial glucose control, which can increase insulin resistance after a meal. He gave a short preview of a closed-loop study to be presented at this year’s ADA (OR-266). The crossover design examined the effect of high and low fat meals on glycemic control. A high fat dinner with identical carb content to a low fat dinner caused more hyperglycemia despite increased insulin coverage by the closed-loop system. He noted the high degree of inter-individual variability as well – one patient needed twice as much insulin for a high fat meal with an identical carb load to a low fat meal. With this in mind, Dr. Wolpert recommends obtaining a diet history and focusing specifically on fat intake. Foods like peanut butter and cheese – traditionally characterized as “free foods” for type 1s – “can cause big problems.”
  • “Setting alarm thresholds on the sensor are like setting basal rates on the pump.” Dr. Wolpert emphasized that the goal is to “derive benefit, but reduce the risk for alarm/sensor burn-out.” When patients first start on CGM, he called for setting conservative initial alarm thresholds: 55-60 mg/dl for lows and 250 mg/dl or more for highs (80 mg/dl or higher as a low alarm for those with severe hypoglycemia or hypoglycemia unawareness). This step should be followed by further tightening and refinement as necessary. Key questions to remember at follow- up include: Did the alarm alert the patient of all low and high glucoses? Did the patient hear/feel the alarm? How many false alarms is the patient experiencing?
Optimizing High Alarm Threshold Optimizing Low Alarm Threshold

If fasting glucoses are often high, but no overnight high alarm:

 

(1) Check whether the patient hears/feel the alarm

(2) Decrease the High alarm threshold.

If frequent hypos but no low alarm:

 

(1) Check whether the patient hears/feel the alarm

(2) Increase the Low alarm threshold.

If the High alarm threshold is going off too often and frequently when glucose is not high:

 

(1) Increase the High alarm threshold or snooze duration to minimize repeat alarms.

If the Low alarm threshold is going off too often and frequently when glucose is not low:

 

(1) Decrease the Low alarm threshold

 

Questions and Answers

Q: Do you recommend a 50/50 split between basal and bolus?

A: That’s somewhat of a generalization. It’s valuable as a starting point, but patients are individuals and you need to get a sense of where a particular problem area might be. Some patients are highly sensitive in the liver and their basal requirements are quite low. I don’t look at it as a treatment endpoint at all. It’s a starting point for defining where problem areas are.

Q: Is there a way on the report to see how long the temp basal or suspend is? Or how many are being done?

A: On the new version of the Medtronic Professional, there’s a column that outlines how many times a person is suspending. Otherwise, you need to look at it on a day-by-day basis.

Q: Do you recommend suspension of the pump for exercise?

A: In practice, that’s a nice feature of pumps. It gets around the issue of weight gain and is one of the advantages of pump therapy vs. MDI – people can reduce their caloric intake around exercise. But you must consider the type of exercise. Isometric exercise gives a big epinephrine surge, and you don’t need a reduction. It’s mainly for just aerobic exercise. With insulin pumps, you have a depot of insulin when you suspend insulin delivery. It can take 30-60 minutes for insulin levels to decline. In those without diabetes, insulin drops very quickly. While you can suspend the basal with pump therapy, to mimic normal physiology is quite complicated. Patients must suspend their basal 30-60 minutes ahead of exercise. The other side of suspending pumps with exercise is the post-exercise period. Some patients rebound up quite high from unrestrained hepatic glucose production. They might need a mini bolus right after exercise. The two main clinical issues are then early basal suspension and to what extent they need a bolus after.

Q: With the fat, is it a combination of duration and total insulin?

A: That’s what the data suggest. My co-investigator Gary Steil is modeling the data and coming up with an optimized bolus delivery pattern. People need more insulin. The confound is there are different effects of different fats. Saturated fats are more of a culprit here. But with some patients eating salad dressings with vegetable oil, we still see this late postprandial hyperglycemia.

Q: What about protein?

A: The literature on that is somewhat mixed in terms of protein’s effect on glucose control. Mechanistically, the amino acids could trigger glucagon release. Or there are more gluconeogenic acid substrates. We didn’t formally study that.

Q: Are there clinical studies on the type of bolus for high fat meals?

A: There have been studies on coverage of pizza and the type of bolus delivery pattern. That was more relating to delayed gastric emptying. There are no studies looking at bolus amounts. Diabetes Tech and Therapeutics reported an article on bolusing that incorporates fat and protein. It was out of Poland. But it was not a crossover study design. The practical confound is inter-individual variability.

Symposium: Continuous Glucose Monitoring – Practical Aspects

COST AND COVERAGE ISSUES

Michael J. O'Grady, PhD (NORC, Bethesda, MD)

Dr. O’Grady gave a very valuable talk on CGM reimbursement, brilliantly integrating clinical knowledge with clear expertise in cost effectiveness. He explained the two major cost effectiveness analyses ($/QALY and budgets), noted the levels established in the JDRF CGM Trial, and showed how payer coverage policies are generally based on that data. One major theme from his talk was that not all CGM manufacturers are created equal, and those who can drive down costs (especially with longer sensor wear) will have stronger leverage with insurers. He also noted that “we’re moving from the clinical side to the cost side” with CGM – usually, this move of medical devices is accompanied by lower efficacy in the real world and thus worst cost effectiveness. However, Dr. O’Grady, believes CGM may be the opposite – if sensors can get down to a cost of two fingersticks per day, they would be cost saving. In that case, “Insurance companies will hug and kiss you on the cheek.” Certainly something to hope for as the new generations of sensors are being developed.

  • Major coverage decisions from payers have generally tracked data from the JDRF CGM trial. Aetna and United cover all patients with type 1 diabetes >25 years and those <25 years with recurrent severe hypoglycemia. CIGNA covers type 1s not achieving optimal control or experiencing hypoglycemia unawareness. He highlighted that these are three of the five largest payers in the US. Medicare does not cover CGM for type 1s, though Dr. O’Grady explained that appeals with lots of documentation can be successful.
  • “Not all manufacturers are created equal.” Dr. O’Grady reviewed the different thresholds for the manufacturers shown below, explaining that companies who bring technologies at a lower cost and similar efficacy will have more pull with payors. “If you’re Aetna, you start the discussion with manufacturer #3,” although that “doesn’t mean you don’t cover #1 and #2.” Dr. O’Grady suggested later in the presentation that longer sensor wear drives costs down, so we would broadly speculate that #1 is Abbott (five-day wear), #2 is Medtronic (three-day wear), and #3 is Dexcom (seven-day wear) in the below table. Dr. O’Grady also pointed out that the mean cost effectiveness ratio is just barely less than $100,000 per QALY (the rule-of-thumb threshold in the US and far below the NICE criteria in the UK of £20,000/QALY ($30,600).

From the CGM Cost Effectiveness Analysis

 

Manufacturer

#1

Manufacturer

#2

Manufacturer

#3

Mean

Total Daily Cost

$13.85

$16.71

$9.89

$13.48

 

Incremental Cost Effectiveness Ratios from the JDRF CGM Trial
 

Manufacturer

#1

Manufacturer

#2

Manufacturer

#3

Mean

Cohort #1

 

(A1c >7% and age

>25)

 

 

$100,970

 

 

$130,060

 

 

$68,051

 

 

$98,679

Cohort #2

 

(A1c <7%)

 

$78,318

 

$97,026

 

$57,170

 

$78,943

  • Insurers ~85% share of the cost of CGM suggests manufacturer discounts may sway coverage decisions. Dr. O’Grady showed a slide with the typical CGM claims for a 90-day supply of sensors The billed amount is $1,380 and insurers receive a 39% discount over retail. Of the remaining $840, the insurance company picks up 85% of the cost ($714) and the patient pays 15% ($126). According Dr. O’Grady, “You can see that this is going to get their attention.” He suggested that the discounts offered by manufacturers may drive payers to choose one manufacturer over another.
  • The length of sensor wear “makes a tremendous difference in cost.” Dr. O’Grady showed an analysis using data from manufacturer #3 in the aforementioned table. By switching the sensor site every seven days, it costs $9.89 per day for a 90-day supply. However, switching the site every 14 days brings it down to $5.38 per day. He emphasized repeatedly that that was just based on anecdotal reports and he is not publishing this data. As a reminder, Dexcom noted in their 1Q12 earnings call that they will pursuing extended durability claims for the G4 sensor (see page five of our report at http://www.closeconcerns.com/knowledgebase/r/109d0417). The current version is for seven-day wear. Medtronic’s new Enlite sensor is six-day wear. We’re not sure about Medtronic’s plans for longer durability, though we assume they are working on it.
  • Dr. O’Grady believes CGM has the potential to be cost saving in the long run. He showed a comparison between the incremental cost effectiveness ratios of two fingersticks per day (ratios of -$1,494 and -$15,725) and use of CGM according to the label ($98,679 and $78,943 from the mean values in column three of the bottom table above). Two fingersticks per day were actually cost saving, suggesting that if they are inexpensive enough, sensors could be too. Dr. O’Grady emphasized that it’s “very rare” to see cost saving interventions and this would be very powerful and persuasive for payors and Congress.
  • “There are reasons to believe that CGM may be more cost effective than in the peer reviewed literature” – we need different analyses moving forward. Dr. O’Grady believes we need more real-world evidence of CGM’s benefits beyond anecdotal accounts. He was especially in favor of claims analyses to track patient use.
  • The two major cost effectiveness analyses (cost per QALY and budgets) help give decision makers a better idea of the return on investment for a particular intervention. Dr. O’Grady explained that in the US, the general threshold for reimbursement is $100,000 per quality adjusted life year, while the “Brits are the hardest” at £20,000/QALY for the NHS ($30,600). The other criteria typically used by the federal government is budget and cost estimates. Such analyses project a spending stream under a current and proposed alternative. Notably, quality of life is not considered in these analyses and they are typically ten-year estimates. Dr. O’Grady explained that this is problematic for a disease like diabetes, where offsetting savings from avoiding complications is probably delayed beyond ten years. Medicare estimates go out 75 years, which he believe should be taken with “a grain of salt.”

Questions and Answers

Q: In the UK, there are a few cases where you can get NHS reimbursement for CGM. It’s for recurrent hypoglycemia. There is flexibility but it’s not great.

Q: Thanks for your presentation. I’d like to comment on coverage by payers. Many of the payer in the US are covering people with type 2 diabetes on insulin. Your slide on coverage policies reflected only type 1s. Many payers are moving into the type 2 world.

A: Good.

Q: Has industry collected data on using sensors for 10-14 days?

A: I don’t know. That’s a hard one. You can see anecdotal evidence, but we haven’t tested this. It is testable. People are clearly doing it and not noticing a deterioration in accuracy that’s encouraging them to switch the sensor. We know our friends at FDA want 100% accuracy all the time and then you swap out. I don’t know if they’ve talked to CMS, who must pay for this.

Q: Nice presentation, thank you. Could you educate us on the reasoning for the lack of coverage of CGM by Medicare? In this day and age, more and more type 1s are in the Medicare population. We know from the T1D Exchange that it’s adults who seem to have a high prevalence and occurrence of severe hypoglycemia. My mother is 81 and has horrible, brittle diabetes. She was hospitalized for days at a high cost. This could have been prevented by a CGM, which she now wears. What do you do in these cases? Is it possible to appeal and get coverage in cases like this? There is a clear documented need. Why is Medicare not covering this when the data suggests otherwise?

A: Our fault as investigators is part of it. In the JDRF CGM Trial, we didn’t have enough seniors. For something like a sample size of 20, they are not going to make a national coverage decision. We could have done better at providing data. But they’re still behind the times. They don’t think many type 1s make it to Medicare age. They just have to run the claims. You see how old people are, you see complications, and you can see how many are billed to type 1 and type 2. The claims indicate, on the actuary side of it, that there are hundreds of thousands of type 1s that are Medicare eligible. They know this. But that’s not on the coverage side though. That is a real sort of education campaign. This is not just 200 people. It’s really hundreds of thousands of people. As the boomers retire, and god knows they’re aware of this, you’ll see more and more every year. You also have people in very tight in control who are 70 years old. Sometimes they have a hard time loosening up on the tight control. Of course they don’t have to worry about blindness in 20 years when they’re 80 years old.

INTERPRETING AND APPLYING THE DOWNLOAD

Bruce Bode, MD (Atlanta Diabetes Associates, Atlanta, GA)

Dr. Bruce Bode gave an outstanding case-based presentation on interpreting CGM downloads and getting paid to do so. He advocates using a macro-micro approach, first using the modal report (14 days or less) to look for broad trends and then looking at individual days. In terms of new CGM users, he finds that about one-third of people figure out CGM on their own and really like it, another one-third really need help and guidance, and another one-third just stop using it. Dr. Bode also ran through the reimbursement criteria to interpret CGM downloads and strongly encouraged the audience to get paid to do it. He concluded with six excellent case studies using the macro-micro approach, which really made download interpretation look simple, fast, and led to some impressive changes for patients in terms of hypoglycemia, A1c, daily insulin doses, and weight. The most common themes were using CGM to identify lows throughout the day and highs after meals. It was really clear that the combination of insulin delivery and CGM data is extremely helpful in this process, and we certainly look forward to more sensor augmented pumps in the next couple years. Dr. Bode advocates taking a macro-micro approach to interpreting CGM downloads. The macro view looks at the big picture to diagnose problem areas. Dr. Bode recommends using the modal day (14 days or less), meal overlay curves (“are they covering meals?”), and overnight curves to take a 30,000-foot approach. On the micro front, Dr. Bode delves into individual days, meals, and other events. These are especially helpful to provide patients with teachable moments. Dr. Bode highlighted the “very valuable” importance of wear time statistics (“If you wear it, you usually do well”).

  • When it comes to CGM, “You must try to get paid for it. We do have codes.” Dr. Bode reviewed the reimbursement criteria for CGM, noting the limitation that an MD, DO, NP, or PA must interpret the download to get paid for it. He recommends having a CDE do the initial interpretation and then having an advanced HCP review it, sign off on it, and generate a bill. Dr. Bode also recommended speaking directly with local carriers.
  • Interpretation of CGM download (code: 95251) – The report must be generated and interpreted, it does not require face to face contact, and it requires an MD, DO, NP, or PA.
  • Technical component (code: 95250) – Includes placement of sensor and training, does not require high a high level HCP.
  • CGM is generally accepted under the following codes: diabetes out of control (250.02, 250.03), diabetes out of control with hypoglycemia (250.82, 250.83), pregnancy out of control (640.80-84), and insulin pump adjustment (V45.85).
  • A number of themes emerged from Dr. Bode’s concluding download interpretations from six different case studies. Using the aforementioned approach, Dr. Bode usually interprets CGM download reports in just five to ten minutes. Three of the cases involved hypoglycemia, which was easily identified on the modal day report. Dr. Bode either lowered basal rates or reduced meal coverage depending on when the lows were occurring. Highs after meals were also quite easy to spot on the modal day report (especially on the Medtronic reports because insulin delivery and bolus calculator data was included) – Dr. Bode recommended increasing mealtime insulin doses (though in one case, doing so required a consequent drop in a patient’s basal), carb counting, and pre-meal bolusing. The micro, daily view was particularly helpful for drilling down into these meal issues. One interesting type 1 patient, a 57 year-old type 1 with a BMI of 36 kg/m2, had unexplained hyperglycemia in the morning. The reason was actually caused by sleep apnea and was subsequently diagnosed and corrected. Overall, A1c typically improved by ~0.5% in most cases shown, especially solid results given reductions in hypoglycemia.

Questions and Answers

Q: You are an expert in interpreting downloads. What about someone that’s new to this?

A: If someone is new, they should come back to see a specialist. I will also spend longer with new patients and teach them how to upload at home. We encourage everyone to upload at home. Looking out to CDEs, if you don’t bill for it, you’ll never get paid. Insurance companies love it when you don’t bill, but you’ve got to bill them to teach them to pay for this.

Q: Do you ever use blinded CGM? And do you use dual wave boluses?

A: I’m more of a believer in real-time CGM. I do use professional real time and we bill accordingly for that. We use masked CGM all the time in trials. We have multiple ongoing trials with masked CGM. There are pros and cons to both. With masked, you don’t have to teach anything. But when a patient sees and experiences CGM, they can come back and learn. One-third can figure out on their own. Another one- third need your help. Dual wave boluses are easy to see on CGM. After eating, blood sugar is normal or going low. Then, two to three hours later, it’s going high due to high fat. We recommend adding more insulin to the meal and covering it over another 2 hours. A split of 60/40 or 70/30. There’s not a lot of science here.

Q: Most of your downloads were in type 1s. Any key clinical points for type 2s?

A: We’ve done a lot of downloads. I’m doing a study with Dr. John Buse on type 2 diabetes and downloads. There is little published there. We really think that type 2s are much more predictable. You see a much lower standard deviation and much more predictable curves. We’re hoping when you do type 2 diabetes, you’ll clearly see the problem – is it a fasting problem, a hepatic problem, a bolus problem? Hopefully next year at this time we’ll have data to share. We’ve also used lots of masked CGM in type 2 diabetes studies. There has been a published paper in a single center study from Walter Reed. It looked at CGM in type 2 diabetes in real time. If patients wear it, they do very well and make behavior changes.

Q: How do you tease out basal adjustment and bolus adjustment? You mentioned the three-hour window. Are there other things to use?

A: For basal or bolus, you need a meal marker. Medtronic has the simple to use bolus calculator. Dexcom must be put in to the receiver. Within three hours, I call it a bolus problem. After three hours, and certainly greater than four hours, it’s a basal problem. When making a change in the basal, you must do it two hours before the event. So if you’re going low at three in the morning, you need to make a change at one in the morning.

CGM: CHOOSING THE RIGHT PATIENTS

Larry A. Fox, MD (Nemours Children’s Clinic, Jacksonville, FL)

Dr. Fox reviewed a variety of continuous glucose monitoring papers in different populations, from landmark studies like the JDRF CGM Trial and STAR 3 to more recent research (e.g., in young children and toddlers [Mauras et al., Diabetes Care 2012] and for the purpose of reducing hypoglycemia [Battelino et al., Diabetes Care 2011]). The technology’s potential has been shown potential in both pumpers and MDI users, and in both well-controlled type 1 diabetes (who might get a benefit in beta-cell function) and poorly controlled type 1 diabetes (at least for those already under intensive therapy – the value is less clear in less-motivated patients, a population Dr. Fox and his colleagues are currently studying in an NIH-funded trial). He concluded that adults generally seem to be better candidates than children and that greater frequency and longer duration of use correlate with better results, and he acknowledged that his talk did not even attempt to address a plethora of other important populations (e.g., pregnancy, type 2 diabetes, hypoglycemia unawareness) or to help clinicians identify individual patients that are likely to benefit from CGM (which we had hoped would be a bigger theme, based on the presentation’s title). It’s important to note that all these “rules” will keep changing as the products develop further.

Questions and Answers

Q: You presented a German/Austrian registry study in which people that had been using CGM for less than 30 days had a higher rate of hypoglycemia than the overall population. Might this have been a case of selection bias – perhaps patients were prescribed CGM because they had frequent hypoglycemia?

A: Absolutely – at the end of the day, that study was just a registry review.

PATIENT SELF-MANAGEMENT ALGORITHMS

Rosanna Fiallo-Scharer, MD (University of Colorado Denver, Aurora, CO)

In this ‘CGM 101’-style presentation, Dr. Fiallo-Scharer introduced the audience to an algorithm for helping patients make the most of numerical and especially trend information. During Q&A, she emphasized the importance of basing therapeutic decisions on confirmatory fingersticks rather than sensor readings.

Questions and Answers

Q: Can you give a timeframe for suspending basal rates in hypoglycemia? Patients need to understand that suspension is not a rescue therapy – suspending basal rates won’t take care of it if they are already too low, like 45 mg/dl.

A: But when you wear a pump that is one feature you can use – to suspend the basal rate until glucose becomes stable again. We don’t recommend replacing with a temporary basal rate – the original rate will automatically kick back in once the suspension is over.

Q: How do you interpret data over the next two days after a patient has a hypoglycemic episode?

A: I am not sure what you mean by that. Frequent hypoglycemia can cause patients to become unaware of their lows due to counterregulatory failure. Some patients tend to overtreat lows – I don’t know if that is what you mean. They can get a rebound hyperglycemic event and then have to catch their tails. One beautiful thing about CGM is that they get that feedback and can fix it next time.

Comment: In my experience hypoglycemia results in a lot of overtreatment – patients often aren’t aware that yesterday’s hypoglycemia can cause today’s hyperglycemia.

Q: Can you clarify whether with the algorithm you are having people respond based on sensor glucose or confirmatory blood glucose?

A: Patients were asked to confirm any alarm with a fingerstick and to base treatment on that. The sensors are more useful for looking at trends. Obviously very often the alarms are associated with a real low or high. But if they are going to act based on a number outside the target range they should use a confirmatory blood glucose check.

Symposium: Behavioral Interventions in Routine Clinical Care – What Works?

SUPPORTING AND MAINTAINING CONTINUOUS GLUCOSE MONITORING USE

Timothy T. Wysocki, PhD (Nemours Children’s Clinic, Jacksonville, Florida)

Dr. Wysocki provided an overview of the behavioral aspects underlying CGM use, especially among children and adolescents using the technology. He first reviewed the long list of behavioral barriers to successful CGM use, ranging from unrealistic expectations to youth-parent conflicts about CGM data. He also noted the disconnect between young patients that have grown up in a world where everything works, versus the fact that CGM is still a new technology (“these individuals are immensely intolerant of glitches”) – this was an interesting characterization that we have not heard before and one with which we agree. The remainder of Dr. Wysocki’s presentation described the NIH/NIDDK funded study his team is undertaking (clincaltrials.gov identifier: NCT00945659). The nine-month trial hopes to enroll 150 poorly controlled adolescents (mean A1c is 9.1% in the 97 enrolled thus far), who will be randomized to one of three groups: standard care, CGM, or CGM plus behavior therapy. The behavior therapy is aimed at promoting the benefits of CGM by addressing each adolescent’s unique barriers. This will take the form of motivational interviewing, problem solving training, and communication training. Sessions will guided by a manual and will have a defined structure, although Dr. Wysocki emphasized that families will dictate the content. Data collection is expected to end in November 2013, with A1c as the primary outcome. Additionally, we believe that Medtronic’s new Enlite sensor and Dexcom’s new G4 sensor will significantly improve the CGM experience for many patients. To boot, we are wary of studies that use old technology that are not always characterized as such. To boot, with this trial, we’re very glad to see such smart and committed focus on the behavioral aspects underlying CGM use. In our view, better understanding of this area will be an integral part of improving the technology and climbing up the adoption curve in the years to come.

  • Dr. Wysocki pointed to a variety of behavioral barriers that prevent patients from successful use of CGM:
    • Unrealistic expectations
    • Inconsistent or infrequent use
    • Deficient calibration technique (“garbage in equals garbage out”)
    • Treating without verifying blood glucose level
    • Non-response to trend alarms
    • Disabled alarms
    • Youth-parent conflict about CGM data
    • CGM associated pain or discomfort
    • Hypoglycemia fear or avoidance
    • Device loss or damage (“Including one run over by a mother’s Cadillac Escalade. It didn’t work but the company retrieved the data”)
    • Peer, school, and fashion issues
    • Information overload
  • While we would absolutely agree there are barriers, and while we believe these barriers are driving lower than optimal satisfaction among patients today, it’s also key to point out that this list has improved over time and will continue to do so as long as the therapy is used. If it is not, that is troubling for all patients who would benefit from future versions of CGM.

Questions and Answers

Q: How do you find this intervention integrating into the clinic?

A: We specifically recruited master’s level social workers. We thought they would be more commonly available in ordinary diabetes clinic settings. They are often called upon to do some of the tasks we’re talking about. Our first effort is to demonstrate efficacy. Then, we can talk about dissemination. But like the previous speaker, I’ve had experience teaching these skills to highly experienced individuals and I really struggle. It’s hard to get people to give up things they’re comfortable with. But the majority of pediatric diabetes centers have people like this around. What remains to be seen is (a) is it effective? and (b) can we disseminate it practically.

Q: You mentioned online data collection – are you doing that in-clinic or on a computer?

A: All of these people had to have a computer in order to download the CGM data. About 92% of families have completed at least some questionnaires. About 70% have completed all. And they love it.

Symposium: Glycemic Variability

METHODS OF QUANTIFYING GLYCEMIC VARIABILITY

David Rodbard, MD (Biomedical Informatics Consultants Potomac, Maryland)

Dr. Rodbard gave an excellent overview of the ongoing debate about measuring glycemic variability – his central thesis was that percent coefficient of variation (%CV), which is the standard deviation divided by the mean multiplied by 100 (SD/mean X 100), is the best measure of glycemic variability. Dr. Rodbard provided a variety of arguments to support this assertion: %CV is easily calculated, is independent of A1c and mean glucose, can be measured more precisely than MAGE or CONGA, and is one of the best predictors of hypoglycemia. He also argued in favor of establishing reference ranges for %CV: <32% is excellent, 32-37% is good, 37-42% is fair, and >42% is poor (Rodbard et al., Postgrad Med 2011). These cut-points also align well with other data sets from Dr. Irl Hirsch (%CV <33% is ideal, while %CV >50% is poor) and the CACT1 Study (<34%, 34-40%, 40-46%, and >46%). As the %CV gets below 25% or 20%, Dr. Rodbard noted that there is no hypoglycemia. Encouragingly, %CV is applicable to both SMBG and CGM and people with type 1 or type 2 diabetes. However, specific cutoffs may need to be established for certain populations (early stage type 2s, pregnant women). Dr. Rodbard was negative on MAGE because “it throws out half the data” by only counting upstrokes or downstrokes but not both. He also highlighted the importance of time horizon, as glycemic variability can be measured within day, between day, and overall. He recommends taking the inter- and intra-day standard deviations as well as the SD of daily means. He also showed the Ambulatory Glucose Profile (AGP), noting that our subjective interpretation of a patient’s glycemic control correlates quite well with %CV (As a reminder, the AGP is what the Helmsley Charitable Trust plans to use to standardize glucose reporting; see our report on the recent panel at http://bit.ly/KcWzAd).

Questions and Answers

Q: I have a comment on a simpler measure: range divided by mean. That’s something that clinicians can look at for very small sample numbers. It tells you something about the risk of hypoglycemia. I agree it’s extremely crude, but it’s quick to do in your head.

A: Range is proportional to SD. But it’s affected by outliers. Those outliers may be clinically interesting. Range based on four to five measurements may be unstable. If you have 288 values or 1000 values, it may be more stable. It might be better to use 10th and 90th percentiles to throw out the lowest and highest values.

Q: Do you have data on whether there is a correlation between this measure and outcomes in the general pop or the pregnant population?

A: No sorry I do not.

Q: Do the values need to be different for type 1 diabetes vs. type 2 diabetes?

A: The values for interpretation will be different. In normal subjects, there is about 17% variability, which rises to a %CV of 20-25% for obese, non-diabetic patients with A1cs in the normal range. We need to define the ranges for interpretation for early type 2s and type 1s.

Q: Is there data correlated the quartiles and quality of life measurements?

A: No, I don’t have that. On the earlier question relating to the correlation between %CV and outcome – there was a paper that appeared Diabetic Medicine on coronary calcium. Some of the different types of SD were correlated retrospectively. I believe these measures will be useful for correlation with outcomes.

METHODS OF MINIMIZING GLYCEMIC VARIABILITY IN TYPE 2 DIABETES

Robert Vigersky, MD (Walter Reed National Military Medical Center, Bethesda, MD)

Addressing a nearly full room on Day #5 of ADA 2012, Dr. Vigersky pointed out that the “turnout on this last day of the meeting is a testimony to how important this topic has become.” Most interesting was Dr. Vigersky’s discussion of CGM in type 2s, which included a review of his study published in Diabetes Care earlier this year (see our report at http://bit.ly/zrqLK9). Showing examples from the trial, Dr. Vigersky cautioned that CGM downloads have gaps where no glucose data is collected – in his view, these can give “very spurious variability data unless you do something about it.” His team chose to use an interpolation approach that approximated the lost values based on the pre- and post-gap CGM values. He then discussed the analysis we first saw at Clinical DTM a few months ago (see pages 11-12 of our report at http://bit.ly/Mm64NT), which classifies CGM users in his study based on their observed response to the device. Dr. Vigersky focused on those who learned over time, burned out, or achieved and maintained tight control. Interestingly, the number of receiver screen views was related to each group: those who burned out over time looked at the receiver increasingly less often over the course of the study, while those who achieved tight control increased their number of screen views over the course of the study. The latter group also had Problem Areas in Diabetes (PAID) scores that were low at baseline and remained low throughout the study, potentially offering a window into identifying types 2s likely to benefit from CGM. He concluded that real time CGM can be used as a behavioral modification tool for type 2 diabetes, although the dose, frequency, and patient selection criteria need to be further studied. Dr. Vigersky also covered SMBG in non-insulin using type 2s and argued that use of structured testing (STeP, ROSES, St. Carlos studies) to selectively prescribe medications can improve both A1c and glycemic variability. Finally, he mentioned some nutritional strategies to reduce glycemic variability: eating carbs at the right time of day (lunch was best in one study), eating lower glycemic index foods, and eating vegetables before carb intake (rather than after).

  • Gaps in sensor readings were fairly common in Dr. Vigersky’s study and were dealt with using an interpolation approach. All 47 patients in the CGM study arm experienced gaps in data. Dr. Vigersky and colleagues looked at CGM data in three-day cycles in the middle of the week (we note that participants were wearing the Dexcom Seven in this study for two weeks on, one week off). In all three-day cycles, the total number of gaps of over five minutes was 14,173, representing 6.6% of the data. This translated to an average number of gaps per subject of 302 and a mean gap of 26 minutes. Dr. Vigersky noted that gaps make it difficult to calculate statistics such as MAGE. To fill in the missing data, Dr. Vigersky’s team took the before- and after-gap values and interpolated. In other words, if the CGM read 120 mg/dl, then a ten-minute gap (one missed reading), then 130 mg/dl, the missing point would be 125 mg/dl. (According to management, Dexcom’s G4 has a significantly improved transmitter relative to the Seven Plus. In product’s recently completed pivotal study, the G4 sensor captured 99% of data points. It also has a typical transmission range up to 30 feet and up to 50 feet if it’s in line of sight, compared to a range of five feet cited in the label for the Seven Plus.)
  • Dr. Vigersky characterized patients with an “immediate effect” response pattern as learning over time, burning out, or achieving and maintaining tight control. Thirty-eight of the 47 patients fell into the immediate effect group. Screen views were defined as discrete episodes of viewing the display with one or more minutes between views.
    • Those who learned over time saw an improvement in mean glucose, standard deviation, and MAGE throughout the study. Their mean number of daily screen views was similar throughout the study, averaging 16-17 per day (i.e., one episode of screen viewing every waking hour). Their Problem Areas in Diabetes (PAID) scores decreased from 32 to 25.
    • Those who burned out got worse over time. Their mean blood sugar increased and their screen views decreased over time (from 12 to less than five). This group’s PAID scores did not change. For us, the central question here is one of causality – are the reduced screen views causing the worse outcomes, is frustration with the device causing reduced screen views and therefore worse outcomes, or is there some other factor at work. We think this would be an interesting qualitative question to study among ex-CGM users; better understanding why people quit using the technology should improve it. More importantly, we believe that churn will decline as product ease of use, reliability, accuracy, etc. improve, which is only a matter of time – a “when” not an “if”.
  • Those who achieve and maintained tight control got immediately better and stayed better. Mean blood glucose improved and standard deviation dropped from 20 mg/dl to 14 mg/dl. The group’s screen views also doubled from about 15 to 30 per day by the end of the study. This group’s PAID scores were low at first and did not change much by the end of the study.

PERSPECTIVE ON OUTCOMES

J. Hans DeVries, MD, PhD (Academic Medical Center, Amsterdam, Netherlands)

Dr. DeVries delivered a skeptical and thoroughly cited review of the proposed link between glycemic variability (GV) and harm in diabetes. Two of the field’s foundational findings (a link between GV and microvascular complications in type 1 diabetes, seen in a 1995 analysis of DCCT data; a link between GV and oxidative stress in type 2 diabetes patients not on insulin) have been retracted or weakened upon follow-up analysis. Dr. DeVries proposed that future studies of GV concentrate on mean absolute glucose change (MAG), a measure developed by his group that accounts for the frequency of glucose fluctuation (rather than just the dispersion of values, a la standard deviation). To close he acknowledged that at the very least, unpredictable values are very disturbing to patients and are associated with risk of severe hypoglycemia. He recommended continuous glucose monitoring as a tool for reducing both A1c and variability (though he indicated that any glucose-lowering intervention will tend to address both A1c and variability at the same time since the measurements inherently related).

  • Dr. DeVries said that substantial doubt has been cast on two of the foundational papers linking glycemic variability to harm in diabetes. These papers were the analysis of DCCT suggesting relationship between GV and microvascular complications in type 1 diabetes, independently of A1c (Diabetes Care 1995) and the famous Monnier study of 20 non-insulin- dependent people with type 2 diabetes, which suggested that 74% of variation in oxidative stress could be explained by glucose variability (JAMA 2006). The authors of the former retracted theirpaper after finding that the GV/complications link was a statistical artifact due to erroneous modeling assumptions. As for the latter, when Dr. Monnier expanded the study size to 60 total patients, GV explained only 15% of variability in oxidative stress (Diabetologia 2010). Also problematic for the Monnier findings was Dr. DeVries’ group’s failure to confirm the relationship with a more accurate assay for oxidative stress (Siegelaar JDST 2011). However, Dr. DeVries acknowledged that his study’s population had lower mean A1c (7.0% vs. 9.0% in the Monnier cohort); other analyses suggest that high GV may be harmful only in those with high mean glucose. Epidemiological data on the question are mixed, and post-hoc analysis of the randomized clinical trial HEART-2D fails to support GV’s relevance as a risk factor independently of mean glucose (with which variability is inherently associated, since people with high mean glucose have a wider range over which to fluctuate).

Questions and Answers

Q: How do we go through this maze? Do we need something like ORIGIN, or can we make sense of the data available?

A: What’s available is secondary analysis. This gives us a hint that large ORIGIN-like study may not give definitive answer, since it would be so difficult to design. The APOLLO study compared basal to prandial insulin and didn’t show a difference, but it didn’t measure oxidative stress, and it wasn’t big enough.

Comment: The FLAT-SUGAR study might address this issue. It is not as long-term as ORIGIN, but it will go on for a year – so we will probably have an answer in the next year or two. (Editor’s note: As we understand it, the yearlong FLAT-SUGAR trial is technically a feasibility study to establish whether different therapeutic regimens can separate cohorts of patients by glycemic variability while causing equivalent effects on mean glucose. The independent clinical relevance of GV change would not be assessed until a larger follow-up study.)

Q: It is thought that exposure to hyperglycemia may have a non-linear effect – more pathology is generated by a few high sugars rather than many lower sugars, even if the mean glucose is the same. The theory behind MAG is different – that harm is related to the act of going up and down.

A: Intuitively it makes a lot of sense that a repeated insult is more harmful than a single, long-standing insult. By definition a stable glucose must be better than variable, since nature preserves stable glucose so strongly. But the additional harm of variability (relative to mean glucose) is probably minimal.

GLYCEMIC VARIABILITY IN THE CRITICALLY ILL PATIENT

James S. Krinsley, MD (Stamford Hospital, Stamford, CT)

Dr. Krinsley reviewed the substantial and growing body of evidence that establishes glycemic variability as a risk factor for ICU mortality, independent of hypoglycemia and hyperglycemia. Ongoing research needs include characterizing the risks of GV by subpopulations (e.g., diabetes vs. non- diabetes, medical vs. surgical ICU), and Dr. Krinsley is leading a massive (n~42,600) international observational study that he expects will aid in this effort. (Preliminary analysis suggests that GV is associated with increased mortality risk whether or not patients have diabetes, but the slope of the risk increase is steeper for people without diabetes.) Dr. Krinsley argued that inpatient glucose control must address all three domains of glycemic control – hyperglycemia, hypoglycemia, and glycemic variability and he looked forward to benefits from wider use of new technologies (e.g., CGM, insulin-dose-recommendation software). He also offered the audience some relatively low-tech recommendations from his own clinical experience (e.g., supplement intravenous insulin with twice-daily subcutaneousinjections of basal insulin as a way to increase glycemic stability; use 10% dextrose solution rather than 50% dextrose solution as hypoglycemia rescue therapy, in order to avoid rebound hyperglycemia).

Questions and Answers

Q: Can you comment on the incremental importance of GV as a factor separate from hypoglycemia?

A: The data suggest that these derangements are cumulative – the most recent paper on this is Mackenzie et al., Intensive Care Medicine 2011.

Q: I am a little lost. I think we are completely lacking in causal effect. The only randomized controlled trial data involved targeting a BG so low that no one is going for it anymore, at least not in the US.

A: Yes, this is a challenging issue, and it is impossible to ethically randomize anyone to hyperglycemia or increased glycemic variability. As to the guidelines, I believe that the one-size-fits-all approach is not necessarily correct, even though it is based on the largest randomized controlled trial on the topic. I think that study, NICE-SUGAR, was the end of Chapter One of inpatient glycemic control. With new studies – like the large observational I mentioned – and with new technologies, we will enter Chapter Two.

Q: I am fascinated by your new multinational data looking at the differences among those with diabetes and without. If I am remembering your slide right, at high glycemic variability levels the mortality was less for people with diabetes than people without diabetes. Did you look at the diabetes medications the diabetes patients were on before entering the hospital?

A: This was huge database study, so we didn’t have that information. Diabetes patients might show up in the ICU with a broader variety of comorbidities, but numerous studies show that diabetes is not an independent risk factor.

Q: Have you started using incretins in the ICU as was proposed at a conference in Brussels two years ago?

A: I haven’t, and I’m not aware of any widespread use.

Symposium: Beyond Insulin and A1C in the Management of Pediatric Type 1 and Type 2 Diabetes

IS IT TIME FOR ROUTINE MONITORING OF OTHER MEASURES OF GLYCEMIC CONTROL?

Thomas Danne, MD (Kinderkrankenhaus auf der Bult, Hannover, Germany)

Dr. Danne discussed “Something that I feel quite strongly about: Looking beyond A1c and insulin.” He began by showing how A1c does not tell a complete enough story, which provides rationale for using glycemic variability. Although there are a wide variety of glycemic variability statistics, Dr. Danne prefers using standard deviation – it’s easy to understand and other measures don’t seem to offer benefits. However, there is certainly an allure to having one number, so something like the Glucose Pentagon may be warranted (combining A1c and four other numbers). Although the evidence is mixed on glycemic variability, Dr. Danne believes it is clinically relevant and using CGM can help improve it. He concluded by summarizing studies of the Veo and the DREAM project, explaining that glycemic variability will help us assess if these closed-loop therapies are beneficial. The Joslin Diabetes Center teaches the three pillars of diabetes are insulin, exercise, and diet; similarly for Dr. Danne, the three pillars of glucose management are CGM, A1c, and SMBG. We’re glad to see more and more focus on looking beyond A1c. And in our view (and, Dr. Danne’s, we’re sure), the key will be convincing FDA and other regulatory agencies that this is the way to go.

  • “A1c does not give us any idea about glucose fluctuations…knowledge of glycemic variability is important for adjusting diabetes therapy.” To illustrate the shortsightedness of A1c and the importance of glycemic variability, Dr. Danne showed a patient’s modal day report with a high average blood sugar (translating to a high A1c) and lots of glycemic variability. He then graphically showed how intensifying therapy to lower the average would bring the patient into hypoglycemia. A different patient with the same average but less glycemic variability would be brought into target after therapy was intensified. We found this simple graphical illustration very persuasive.
  • “I still believe standard deviation is very easy to understand and maybe the best parameter. Even though all my math friends tell me it’s the wrong thing to do.” Theoretically speaking, Dr. Danne explained that standard deviation is hard to use because upswings are larger than downswings (i.e., there is a glycemic floor). However, he believes standard deviation does as good of a job as other measures and is much easier to understand and calculate. For a more complete review of other glycemic variability measure, he told audience members to read Dr. Hans DeVries’ paper in Endocrine Reviews (2010). Dr. Danne conceded that We always want one single value,” so something like the Glucose Pentagon may be warranted (Thomas et al., Diabetes Technol Ther 2009). This measure combines A1c, GPR, standard deviation, time spent >160 mg/dl, and AUC >160 mg/dl.
  • Regarding the clinical relevance of glucose variability, Dr. Danne stated, “It’s still controversial. We need more data.” He first reviewed the famous Monnier study linking oxidative stress to MAGE, which has not been duplicated in a number of other studies. To provide some more clinical data, Dr. Danne next discussed an interesting analysis of 1,000 insulin pumpers. The number of boluses per day was linearly associated with A1c – as boluses per day increased from two, four to seven, and then to over 12, A1c declined in a step-wise fashion from 10.4% to 8.2% to 7.3%. He likened this to driving a car and steering it several times – the more times you steer it, the more likely you are to stay on the road. Since this study did not look at glycemic variability, Dr. Danne believes it would be interesting to combine the pump data with CGM. One would expect a high number of daily boluses would lead to a drop in glycemic variability.
    • Dr. Danne also highlighted the ONSET trial (see pages 107-109 of our EASD 2011 Full Report), which compared conventional pump therapy plus SMBG to sensor-augmented pump therapy from the onset of diabetes. Those on SAP therapy had higher C-peptide levels, potentially suggesting that the reduced glycemic variability associated with CGM may have reduced glucose toxicity and helped preserve more beta cells. This was an interesting hypothesis and we hope that the TrialNet Metabolic Control in New Onset Diabetes (clinicaltrials.gov identifier: NCT00891995) study can help answer it. There is a poster on the trial at this year’s ADA (891-P), though the full C-peptide data isn’t expected until November 2013.
  • Studies from Dr. Danne’s group suggest glycemic variability improves with CGM. Dr. Danne discussed two CGM studies using the Abbott Navigator CGM (Danne et al., Diabetologia 2009) and another with the Dexcom Seven Plus. Patients used both a masked and unmasked sensor.  Independent of baseline A1c, patients were able to lower glycemic variability with use of the real-time CGM.

Questions and Answers

Q: I have a small pediatric diabetes program and we have lots of kids on sensors. My problem is they don’t stay on sensors. How do we keep kids using sensor? Also, after listening to all of the closed loop and CGM data for two days, I’m worried about kid’s real estate. We saw a glucagon pump, an insulin pump, and a sensor. How are we going to keep this real estate intact?

A: Imagine that while you were driving your car, you had to insert two needles into your body and you got several alarms all the time – you should have slowed down because of that red light. Or there’s a bicycle over your shoulder. After three days, you’d probably take the bus or walk to work. [Laughter] I’m not surprised many patients are saying, “I’m fed up with the sensors. They’re alarming and not helping me.” It’s a technology issue. Yes, there is a real estate issue. But we are burdening them with information. Patients want simple solutions. Something like LGS, which suspends without an alarm. Or overnight closed loop – you hook onto it in the evening and wake up in the morning with great control. That’s what our patients want. I think the real estate issue will be smaller.

Q: You showed that taking 12 boluses a day was associated with a lower A1c. I actually see that kids bolusing a lot have a lot more glycemic variability. Was there any measure of glycemic variability in that data?

A: I would love to have that data. Taking 12 boluses per day is like having a second basal rate. It simply came out that way. There was a clear relationship between the number of daily boluses and A1c. Seven to ten is a good number. That’s certainly much higher than injection therapy. I would not ask any patient to take ten injections per day.

Symposium: Joint ADA/EASD Symposium – Capillary Glucose Monitoring in 2012

SELF-MONITORING OF BLOOD GLUCOSE IS USEFUL IN PATIENTS WITH TYPE 2 DIABETES MELLITUS ON ORAL AGENTS — PRO

Lutz Heinemann, PhD (Science & Co, Dusseldorf, Germany)

Dr. Lutz Heinemann made a thoughtful defense of SMBG in non-insulin-treated type 2 diabetes (NITT2) patients, including a meta-analysis of 12 meta-analyses on the topic and some ‘meta-’ discourse about the controversy itself. Meta-analyses tend to show a six-month A1c improvement of 0.3%, and recent randomized controlled trials like STeP, ROSES, and St. Carlos have demonstrated benefits of 0.5% (Dr. Heinemann argued that these studies, all conducted after 2008 and all emphasizing SMBG as an educational/motivational tool, are designed better than historical trials – a positive result of pressures from cost-controlling agencies such as Germany’s IQWiG). He believes that the field still needs more long-term data as well as a neutrally funded, conclusive study that clearly quantifies the study effect (i.e., one that includes a placebo group receiving no intervention at all); he proposed that the major SMBG companies could divert 10% of their marketing budgets for a few years toward such a study. He also expressed optimism about phone-connected meters to improve the integration of SMBG into daily life, and – as a bolder way to encourage proper testing – he proposed that SMBG could require a driver- license-like process of training and certification (Heinemann et al., JDST 2012).

 

SELF-MONITORING OF BLOOD GLUCOSE IS USEFUL IN PATIENTS WITH TYPE 2 DIABETES MELLITUS ON ORAL AGENTS — CON

Jeffrey W. Stephens, MBBS, PhD (Swansea University, Swansea, United Kingdom)

Dr. Stephens took the con side of the debate and argued that SMBG does not improve A1c to a clinically significant level in non-insulin treated type 2s. He ran through a series of professional guidelines and recent studies on the topic, concluding that the evidence is not clear on the subject. In RCTs and meta- analyses that did find a benefit of SMBG in non-insulin using type 2s, the average A1c improvement was typically ~0.25% (“We would not approve this if it was a new therapy for type 2 diabetes”). Dr. Stephens also emphasized the high cost of glucose monitoring, the decrease in well-being, and the potential for an increase in depression. In addressing Dr. Bill Polonsky’s STep Study, Dr. Stephens noted that the benefit of structured testing over the control group was only an additional 0.3% benefit – again, not clinically meaningful in his view. He closed, however, by emphasizing that a structured approach to testing that uses the blood glucose data meaningfully may be warranted in certain patients: those who are educated, motivated, and at risk of hypoglycemia, have inter-current illnesses, are fasting, or when using sulfonylureas. (One editorial perspective is this – we believe the traditionally lauded RCT design of most trials hurts BGM, since it is so “not” real-world in our view. We hope for more realistic trials moving forward.)

ACCURACY STANDARDS FOR SELF-MONITORING OF BLOOD GLUCOSE — ARE THEY ATTAINABLE?

George S. Cembrowski, MD, PhD (University of Alberta, Edmonton, Canada)

Patients with pumps and hypoglycemia unawareness, among others, require relatively high accuracy, as do people in the ICU – a particular emphasis of his talk. (As a side note, Dr. Cembrowski said that hospitals often favor central laboratory glucose testing because it is the least expensive option, even though it is also the slowest.) For hospital systems, he favors the CLSI’s proposed targets of within 12 mg/dl for values below 100 mg/dl and within 12.5% for higher values; he indicated that the upcoming new ISO 15197 requirements (95% of results within 15 mg/dl for values below 100 mg/dl or within 15% for higher values) are also a move in the right direction. Briefly reviewing two anonymous hospital-use meters, he concluded that BGM products seem to be accurate enough in the hyperglycemic and upper- normoglycemic ranges, and “probably” in the lower-normoglycemic range as well. He said that the picture is less certain in hypoglycemia, in part because accuracy testing in hypoglycemia is often “contrived” – i.e., based on altered blood samples rather than blood from actual hypoglycemic patients.

PANEL DISCUSSION

Lawrence Blonde, MD (Ochsner Health System, New Orleans, Louisiana); Andrew J.M Boulton, MD (University of Manchester, UK); Lutz Heinemann, PhD (Science & Co, Dusseldorf, Germany); Jeffrey W. Stephens, MBBS, PhD (Swansea University, Swansea, United Kingdom); George S. Cembrowski, MD, PhD (University of Alberta, Edmonton, Canada)

Comment: We have to be careful to look at studies with biases. In the Farmer et al. study, there was no transfer of information to meaningful therapy decisions. You have to look at those studies differently. If you think about SMBG and insulin treatment, there’s no doubt it is of value. The information is directly transferrable to insulin. This principle can also be applied to last year’s STeP study – glucose information was transferred into a medically meaningful modification of drug therapy or lifestyle intervention. The absolute effects were comparable to effect sizes with drugs.

Dr. Stephens: There’s been uncertainty in the previous studies.

Dr. Heinemann: This is clear. The more recent studies are the ones that we should take more into consideration. Meta-analyses have limitations and they cannot be better than the studies they include. I believe in the STeP study and it was an important step in the right direction.

Q: Talking about outcome, there is a fixation on A1c – a measure of average glycemic control. Glucose variability has not been looked at. My second question is about testing sugars in sulfonylurea patients. You said that’s reasonable, but it may not be for non-hypoglycemia causing therapies. Is there any study looking at that specific question?

Dr. Stephens: I’m not aware of any studies. The majority of studies performed were before the era of DPP- 4 inhibitors and GLP-1 analogs. If any study is designed now, it should take into account use of these agents. Perhaps it should have more than one arm.

Q: I would argue in one of studies, it was said that newly diagnosed had a higher reduction in A1c with blood glucose testing. But they were probably on treatments that did not cause hypoglycemia.

Dr. Boulton: In patients using therapies that don’t cause hypoglycemia, testing might be beneficial due to better adherence to lifestyle recommendations.

Comment: Surely the answer is somewhere between yes and no. If I do a lot of testing and find that my sugar goes up after supper, aren’t we better off for that?

Dr. Blonde: Individualization was a big theme of the ADA/EASD position paper, and it seems like this is a key area where therapy can be individualized. The patient and healthcare provider can decide together if SMBG can be used to improve care.

ECONOMIC ASPECTS OF SELF-MONITORED BLOOD GLUCOSE

Philip Clarke, PhD (The University of Melbourne, Melbourne, Australia)

Dr. Clarke told a waning crowd in this afternoon symposium that while it is established that SMBG is beneficial for those with type 1 diabetes, the evidence is less clear for type 2 diabetes. Despite the evident short-term benefit SMBG has on A1c levels in people with type 2 diabetes, there have been mixed results on its long-term benefit in terms of mortality, quality of life, and cost-effectiveness. He explained that while the evident costs of SMBG are quite clear – in 2002, SMBG cost Medicare B nearly half a billion dollars – it is more difficult to define and measure the benefits. For example, Dr. Clarke said that the information provided by SMBG may be positive (e.g., encouraging more exercise) or negative (e.g., making someone more anxious) depending on the circumstances. However, UKPDS did find a reduction in mortality due to SMBG in people with type 2 diabetes, but it took 12 years for the reduction to become significant. The results have been particularly mixed when researchers attempt to determine the impact SMBG has on quality adjusted life years (QALYs; one QALY equals a year of full health and zero QALYs equals dead) of people with type 2 diabetes. Dr. Clarke next turned to costs, arguing that the United States must start thinking about which health care technologies are worth funding and which are too expensive. He concluded by providing several suggestions for alternative ways of measuring the cost- effectiveness of SMBG.

  • The impact of SMBG on the quality of life adjusted years (QALYs) in people with type 2 diabetes has been inconclusive, making it difficult to do a comprehensive cost-benefit analysis. To see if type 2 diabetes patients have better quality of life due to SMBG, Dr. Clarke looked at participants’ quality adjusted life years using the open, parallel group, randomized Diabetes Glycemic Education and Monitoring trial) DiGEM. During the trial, QALYs were actually reduced in the treatment arms receiving less intensive SMBG (-0.008 QALYs) and more intensive SMBG (-0.035 QALYs) than in the usual treatment arm. Dr. Clarke said the reduction in QALY was mainly due to anxiety and depression. In contrast, the Center for Outcomes Research (CORE) assumed effectiveness based on effects observed in the Kaiser Permanente diabetes registry. This registry included 5,867 patients who were newly beginning SMBG and were only on oral anti-diabetic medications. They found that an A1c reduction of 0.32% results in an additional 0.103 QALYs. Cameron et al. (CMAJ 2010) found similar results: an A1c reduction of 0.25% (baseline not reported) translates to 0.024 QALYs gained.
  • Dr. Clarke said that given the current state of healthcare spending, the United States must start making decisions about which therapies and treatments are cost effective. In his view, disinvestment from ineffective therapies needs to be a critical component of efforts to control healthcare expenditures in the United States. He emphasized that when it comes to health care systems, early interventions are generally going to be easier than waiting until a crisis hits and only hard choices are left on the table. Thus, there needs to be a definitive study on the cost effectiveness of SMBG in treating type 2 diabetes. One method he proposed for such a study would be to assign some predetermined target as effective (for example, an A1c reduction of 0.5% in three to five years, with no reduction in quality of life). If SMBG demonstrated that it was able to meet this target then it should be kept; otherwise, it should not. Such a study, of course, would be expensive, and coming to a consensus on what is effective would be challenging as well. That said, such a study, especially if it incorporated CGM and regular tests on whether therapy was “working” might help us move toward more consensus on what personalized and individualized medicine means.
  • Dr. Clarke proposed trying to lower the costs associated with SMBG by researching if a cheaper test-strip is equally effective, or by motivating meter and strip manufacturers to help people with diabetes see the benefits of better outcomes. He proposed a system in which one would only pay the manufacturer a portion of the meter and strips’ costs initially and then only pay the rest if the individual’s levels stayed below a wanted target. We think that while making diabetes management more cost-effective is critical, there are too many variables to make this suggestion practical – some patients, for example, have the funds to eat good and healthy food while others do not; some have jobs and families that prompt stress and in turn cause diabetes management to be more difficult than others – etc! We do appreciate critical thinking on the part of Dr. Clarke, however, and hope that his talk spawns other suggestions of note.

Symposium: Which Technologies Have Impact on Clinical and Behavioral Outcomes in Diabetes?

BEHAVIORAL AND CLINICAL OUTCOMES OF CONTINUOUS GLUCOSE MONITORING – WHAT IS THE EVIDENCE AFTER A DECADE OF CONTINUOUS GLUCOSE MONITORING?

John Pickup, MD, PhD (Kings College London School of Medicine, London, UK)

Dr. Pickup gave a very comprehensive literature review of the clinical and behavioral evidence supporting use of CGM. He began by showing a number of studies to suggest that A1c improves with CGM, but only if patients wear the sensor (after showing the famous summary slide of the JDRF CGM trial, Dr. Pickup humorously quipped, “I’ve seen this slide four times today. I’m bored of my own talk”). In addition to sensor usage, his 2011 meta-analysis also found that baseline A1c and age were related to the magnitude of A1c benefit. Dr. Pickup asserted that there is good RCT evidence to support improvements in glycemic variability and mild/moderate hypoglycemia with CGM; however, there is no RCT evidence to suggest that CGM improves severe hypoglycemia – trials have not been designed or powered to test this, they had very low levels of severe hypoglycemia at baseline, and did not test CGM specifically for those with disabling hypoglycemia. He showed that there is data to support use of CGM in special circumstances like pregnancy, type 2 diabetes (“an emerging application for CGM”), and in the hospital (good accuracy with current devices). Although observational studies have seen a positive effect of CGM on behavioral outcomes, “surprisingly and disappointingly,” RCTs have not shown a benefit. Dr. Pickup believes insensitive psychosocial measures, high baseline quality of life, and low baseline hypoglycemia and A1c may play a role. Turning to frustrations, Dr. Pickup highlighted that good coping skills (stoicism, problem solving), good use of information (trend and pattern recognition), and significant other involvement are important for success with CGM. Concluding, he stated that all behavioral and quality of life outcomes need to be studied in the target groups where CGM is indicated (continued disabling hypoglycemia and continued high A1c).

Questions and Answers

Q: Your individual patient meta-analysis was an extremely interesting application. Have you thought to use the same techniques in analyzing patient reported behavioral outcomes? Aggregated effects on patient reported outcomes may show little effect, whereas certain patients may have whopping effects.

A: The main reason we did not look at behavioral outcomes is we didn’t have that data. We had to write to all the study investigators and tried to get hypoglycemia measures and A1c. That was a moderate struggle in its own right. It would be good to move on to behavioral outcomes next. That’s a very good idea. I agree with you on the danger of looking at mean values.

Q: Most of the studies you referenced used many products and sensors that are no longer in use. Benefits from CGM are dependent on patients using devices, trusting the information, and acting on the information. There are significant differences between products and across generations of products. A meta-analysis does not give an accurate depiction of the way CGM is.

A: That’s a fair comment. That’s a constant complaint of meta-analyses – they look at the past and don’t capture the present. You have to wait until there is a significant number of trials before doing a meta- analysis. But the message remains the same. Sensor usage and baseline A1c and the reduction of mild to moderate hypoglycemia. What changes is the accuracy – maybe the magnitude of A1c effect through increased training programs and learning to use it. All your points are fair.

Q: There is a group of patients that frustrates our team: we think they’re good candidates, they meet all the criteria, and yet they stop using this device. And it stops rather quickly, within two weeks or two months. And they drop way off. How do we do that way better?

A: It’s the same with pump therapy. There are frustrated patients that you think ought to do well on pumps. The same applies on CGM. There are behavioral issues. We need to learn much more about that. This is an improving technology; it’s not a mature technology. Think of the early days of the computer. They were a blooming nuisance to work with. Dramatic improvements in CGM is my hope for the future so we can help patients like that.

TECHNOLOGY AND YOUTH – HOW CAN CHILDREN, ADOLESCENTS, AND EMERGING ADULTS BENEFIT FROM NEW TECHNOLOGY?

Korey H. Hood, PhD (University of California, San Francisco, San Francisco, CA)

Dr. Hood based his talk on a model of successful diabetes technology for pediatric and young adult patients. He explained that many of these patients lack important diabetes knowledge and skills, many place diabetes as a relatively low priority in their lives, and many are tech-savvy and have short attention spans. Dr. Hood thus believes that in general, the best effects on patient outcomes will be brought about by solid, visually engaging technologies that reduce the burden of care and act as a “scaffolding” to help people improve their self-management skills. The most straightforward tools are “direct,” such as historical insulin pumps and glucose meters. “Direct-plus” technologies, like continuous glucose monitoring, are those that add a layer of pattern management or trend analysis. Dr. Hood said that such direct-plus tools have been shown to simplify pattern management and to improve glycemic control, but their effects on self-management skills and burden of care are still not well characterized. As for “facilitator” technologies such as mobile applications, Dr. Hood noted that a variety of products are user-friendly and have shown promising early results. However, he remains unsure about whether the use of these tools in their present forms is sustainable.

Questions and Answers

Q: Do you have some indications of which technologies are which useful for most patients? Is it consumer-driven?

A: I think there is a lot of self-selection, which suggests that we should more frequently include more youth and young adults with diabetes in these decisions.

Q: I am also very tech-savvy and would love to use systems to facilitate diabetes control. But we are inundated with new technology every three-to-six months – how will we know which tools and websites are effective and worthwhile?

A: I don’t think there is an easy answer. At the pace of technological development, we cannot keep up – apps do not need to go through the FDA, so they can be developed quickly. I think a good strategy is for find apps that complement their philosophy of care and then recommend those (rather than reactively investigating tools that patients ask them about).

Product Theaters

IBGSTAR INNOVATION, INTEGRATION, INSPIRATION (SPONSORED BY SANOFI)

Bruce Bode, MD (Emory University, Atlanta, GA)

After providing an overview of the history and importance of blood glucose meters, Dr. Bode discussed the new iBGStar meter, the first meter integrated with Apple’s iPhone and iPod touch (for our first take on the device and marketing strategy, see our report at http://bit.ly/Jikjhd). Before launching into the specific benefits of the device’s smartphone integration, Dr. Bode mentioned other strong points of the device, including that it is currently the smallest blood glucose meter in the world, it requires a small (0.5 microliter) blood sample size (slightly larger than the 0.3 microliters required for Abbott’s FreeStyle strips), and it provides results within only six seconds. Dr. Bode also touted the device’s accuracy, which he attributed to its dynamic electrochemistry – as we understand it, this innovation makes the iBGStar better able to compensate for environmental, blood sample, and manufacturing variations than the static electrochemistry used in other meters. He then detailed how the iBGStar seamlessly integrates with the iPhone and iPod touch, allowing users to easily display, manage, and share their diabetes information. Dr. Bode also emphasized the intuitive, user-friendly nature of the interface, which in his view doesn’t even need an instruction manual to use. Turning to higher-level functions, Dr. Bode mentioned the meter’s ability to analyze blood glucose data, especially for identifying above, below, and in target range values over customizable time periods. He noted that the iBGStar automatically generates a logbook – these readings cannot be changed, but they can be tagged with additional notes and information. Dr. Bode further emphasized the iBGStar’s data sharing potential, as the iBGStar can automatically generate and email reports and spreadsheets to the user’s healthcare provider or whoever else he or she chooses. He closed by sharing a number of anecdotes and a case study to help illustrate the high potential utility of the iBGStar both as a more discreet and attractive blood glucose meter and as a way for people with diabetes to more effectively monitor and manage their glucose as part of their treatment program.

  • Dr. Bode explained that smartphone integration provides a way for patients to significantly increase their access to monitoring technology without even realizing it. Dr. Bode quoted a statistic that 60% of people with type 1 diabetes leave home without a needle, but he then suggested that virtually none of them go out without their cell phone. There are 331 million total cell phones in the United States – greater than the total US population of 313 million – of which 165 million are smartphones. Dr. Bode noted that Apple in particular has enjoyed exponential growth with its iPhone, growing from 2.1 million in 2008, to 6.4 million in 2009, and then ballooning to 53 million in February 2012. He estimated that about 1.6 million people with diabetes in the United States have either an iPhone or an iPod touch.
  • The iPhone and iPod touch provide several features that improve the iBGStar’s efficacy as a blood glucose meter. Dr. Bode explained that Sanofi sought a partnership with Apple partly because, no matter how the design of the iPhone or iPod touch may change, the actual USB port is the same for all devices and never changes. This gave them the confidence that the product would be usable and useful long-term.
    • He noted its ability to share data both with family members (see below) and with health care providers. Dr. Bode noted that if the vast majority of HCPs do not currently download and look at patients’ blood glucose data, the iBGStar helps remove one step by sending them all data directly (of course, only time will tell if this will overload HCPs and/or how easy reimbursement is).
    • Dr. Bode also explained that the Diabetes Manager App could be constantly tweaked and improved based on feedback both from patients and HCPs, and then existing users could simply upgrade the app automatically. The result is greater utility of the iBGStar to change and evolve than its counterparts. Indeed, many features iPhone users take for granted are of great use in blood glucose monitoring – as Dr. Bode noted, because the iPhone’s internal clock constantly syncs with satellites, data entry on the iBGStar will always have the correct time, something that is far from guaranteed in other meters.
  • The iBGStar is particularly useful in helping children with diabetes and their parents monitor blood sugar. Dr. Bode noted that most children nowadays barely ever put their cell phones down, which among other things means that parents of children with diabetes can rest easy that their children always have their meters close at hand. Moreover, the ability toshare data between Apple devices means that parents can receive instant updates on their children’s blood glucose levels even when they are at school. He also remarked that parents don’t have to give a cell phone to their younger children with diabetes to use the product – Dr. Bode mentioned that he advised the mother of a five-year-old with diabetes, who was understandably wary of giving such a young child an iPhone, that the child could use an iPod touch instead and still enjoy all the functionality of the iPhone version as long as there was wireless access. Dr. Bode also shared the amusing but heartwarming anecdote of how a teenager, recently diagnosed with type 1 diabetes and feeling quite down, was informed that he would be getting his own iPhone so that he could use the iBGStar and immediately declared, “This is the best thing that has ever happened!”
  • People with diabetes have responded very favorably to the iBGStar, both in terms of reactions to the product itself and in terms of improved diabetes management. Dr. Bode shared a number of anecdotes and case studies meant to illustrate the iBGStar’s appeal. He remarked that during testing, when people were first shown the iBGStar, three out of five people pulled out their current meter and said, “You mean I don’t have to carry this around anymore?” He went into some detail with a case study of a 57-year-old man with type 2 diabetes who in February 2012 had a blood glucose of 358 mg/dl and an A1c of 10.9% who switched from metformin, SUs, and a GLP-1 to a metformin- and insulin-based regimen in conjunction with frequent blood glucose monitoring. The patient had specifically mentioned his concerns that he would not be able to regularly monitor his blood sugar due to his heavy work schedule, but he did already have an iPhone and thus could use the iBGStar. Dr. Bode then showed the patient’s data as gathered by the iBGStar – weekly and monthly glucose values and clear indications of his time spent in and out of his target range. The patient, now on a heavy 100 units per day basal dose of insulin, was able to meet his three-month glycemic target and bring his A1c down into the 6% range.

Questions and Answers

Q: Is this an exclusive agreement with Apple? Is this the only device that will be allowed to connect to Apple phones?

A: That’s up to Apple and who they take. I presume others will probably do it. But Apple is very picky. They have to accept you – you don’t accept them.

Q: Could there be an additional app to manage insulin pump settings?

A: Here’s the problem. When you get into a consumer device managing a medical device, you can’t do it. It can transmit data, so you can technically take your blood sugar and put it on your phone, as long as all the calculations are done in the transmitter. That’s what happening right now. You calculate in the transmitter and submit it to the phone.

HELPING PATIENTS MANAGE PATTERNS OF HIGHS AND LOWS BETWEEN OFFICE VISITS (SPONSORED BY LIFESCAN AND ANIMAS)

Steven Edelman, MD (University of California San Diego Medical Center, San Diego, CA)

Given how little time people with diabetes are able to spend with their HCPs, Dr. Edelman highlighted the importance of empowering individuals to independently manage their blood glucose. He explained that LifeScan’s new OneTouch Verio IQ blood glucose monitor, which features PatternAlert technology, help solve many of the challenges faced by patients monitoring their blood glucose. Dr. Edelman, who uses the device himself, also focused on the product’s ease of use, the accuracy of the OneTouch Verio Gold Test Strips, and the device’s eco-friendly design. Ultimately, Dr. Edelman presented the Verio IQ as a way to optimize insulin therapy and stimulate insightful conversations to facilitate better self- management. To see our review of the meter in diaTribe, please visit http://diatribe.us/issues/41/test- drive.

  • Pattern management leads to improved glycemic control, but most patients act in the moment rather than making decisions based on patterns. Dr. Edelman cited a study demonstrating a 2.6 times higher chance of severe hypoglycemia following clinical patterns of low blood glucose (Lee-Davey J. et al, LifeScan Scotland). Unfortunately, in a study of 315 individuals using insulin, 76-79% treat out-of-range blood glucose results immediately, whereas only 10-17% review other recent out of range results to see if there if there is a connection, trend or pattern. (We would also note that making decisions based on patterns is challenging – historically, software has not been designed to do this and few patients actually download their meters.)
  • The OneTouch Verio IQ automatically provides high and low blood glucose pattern alerts. The meter analyzes past blood glucose data as soon as it is measured. For example, if a user has two lows within the same three-hour period over the previous five days, the Verio IQ will immediately alert the user indicating a low pattern with the message, ‘Looks like your glucose has been running low around this time.’ Upon clicking on the alert, the patient can see the specific glucose values and times of the day used by the meter to identify the pattern.
  • People with diabetes often don’t understand the cause or solution to high and low glucose patterns; the OneTouch Verio IQ Pattern Guide provides insights into their causes and offers potential guidance. Though this paper guidebook is not meant to replace the advice of a physician, Dr. Edelman believes that it can be a powerful informational tool for patients to learn how to be proactive when interpreting blood glucose data. One side of the guidebook addresses high patterns, while the other side deals with low patterns. The guide includes pullout tabs for different periods of the day (before breakfast, after dinner, etc.) and suggests potential causes and reasons why the pattern may have occurred. For instance, on the side for a low pattern in the pullout for “Before Breakfast,” potential causes include “Long periods of increased activity” and “Too much intermediate or long-acting insulin before bed.” It also lists potential actions to take, such as adding a bedtime snack or reducing insulin doses. We like the user-friendliness of this guide and are glad to see LifeScan easing the often-challenging connection between blood glucose values and therapeutic changes.
  • Dr. Edelman listed other innovations of the OneTouch Verio IQ, emphasizing its eco-friendly battery and the accuracy of its OneTouch Verio Gold Test Strips. He pointed out the illuminated testing area, high-resolution color screen, and one-step meal tagging to highlight the product’s ease of use. The Verio IQ’s battery is rechargeable (requires charging twice a month based on five tests per day) and has a convenient one-minute rapid charge feature if the power is too low for testing. The Gold Test Strip Smart Scan technology scans 500 times, correcting for common interferences and does not interact with commonly used drugs (except for xylose.)

 

Corporate Symposium: Introducing FreeStyle InsuLinx – The Latest Advancement for Insulin Users (Sponsored by Abbott)

WELCOME AND OBJECTIVES

  • Ralph DeFronzo, MD (University of Texas Health Science Center, San Antonio, TX) Dr. DeFronzo opened by noting the symposium’s focus: introducing the newly FDA-approved Abbott FreeStyle InsuLinx meter (see our take on the recent approval and our first experience with the device at http://bit.ly/I3imrO). Dr. DeFronzo characterized the new meter as “quite an innovative advancement in blood glucose monitoring” and a “unique glucose monitoring device.” He noted that patients face a number of barriers to achieving good glycemic control, and the FreeStyle InsuLinx has been designed to overcome some of these barriers. Before discussing each speaker’s topic and background, Dr. DeFronzo concluded with a reminder that “we still don’t do a great job in the US overall” – over half the people with diabetes in the US have an A1c >7% and 25% have an A1c >8% (a source wasn’t named). Dr. DeFronzo attributed some of this lackluster performance to patients and physicians failing to review blood glucose data. If we can improve patient-physician integration and information gathering (presumably by using the FreeStyle InsuLinx), Dr. DeFronzo asserted that we can “markedly improve the level of glycemic control.” We’ll certainly be interested to see future studies of the FreeStyle InsuLinx to that effect.

FEATURES AND BENEFITS OF FREESTYLE INSULINX SYSTEM

Bruce Bode, MD (Atlanta Diabetes Associates, Atlanta, GA)

In the symposium’s main talk, Dr. Bode gave an upbeat overview of the useful features in the FreeStyle InsuLinx meter. He honed in on the FreeStyle InsuLinx’s Auto-Assist software and emphasized its “very unique” on-meter storage (requiring no downloads or installation), the very useful Snapshot report, the excellent graphical display of paired pre-meal/post-meal testing (“the major benefit of the FreeStyle InsuLinx”), and integration of data with logged insulin doses. Dr. Bode acknowledged that the US version does not have the bolus calculator, saying this was “obviously something to do with FDA”. He asserted that the FreeStyle InsuLinx is designed for insulin users, though he also believes it’s “great for non-insulin users too.” The latter group on diet and oral therapy can especially benefit from the paired testing (reminding us of Roche and Dr. Polonsky’s work on structured testing). Dr. Bode was also a fan of FreeStyle InsuLinx’s testing reminders, its very quick setup, and the simplicity for patients to remotely send PDFs and results for feedback. Dr. Bode closed by presenting three case studies (a type 1, a type 2 on MDI, and a type 2 on basal-only) – the emphasis was on diagnosing problem areas and giving patients feedback. We thought he persuasively highlighted the real clinical value of the FreeStyle Auto-Assist Software in this part of the presentation. Although he was a fan of the touchscreen (“so simple”), Dr. Bode’s features discussion was very clearly centered on the FreeStyle InsuLinx’s software. We’ll be interested to visit Abbott’s booth to further understand how the company is positioning the meter for the US market. 

  • Dr. Bode urged attendees: “The report you must print out is the Snapshot report. You must look at this.” This report includes glucose and logged insulin data. Dr. Bode highlighted the ability to customize target ranges based on patient type (pregnant vs. older with cardiovascular disease risk) and the clear display of mean blood glucose and standard deviation. He recommends that the mean blood glucose divided by two is an acceptable standard deviation. Dr. Bode also emphasized the importance of getting patients to log their insulin doses (on an interesting aside, one of Dr. Jane Seley’s patients has used the FreeStyle InsuLinx to log Symlin doses). The glucose data displayed includes a bar graph depicting time in- and above-target, standard deviation, average blood glucose, test frequency data, trends and averages over time, and notes (e.g., “100% of BG values above target range (70-140 mg/dl) in morning” or “BG standard deviation may not be the best indicator of glycemic control because the average is outside the range of 110-180 mg/dl”). The Snapshot report also displays logged insulin data, including total daily dose and basal-bolus ratio.
  • Dr. Bode concluded the presentation with three case studies using FreeStyle Auto- Assist reports – themes included the ease of diagnosing problems and changing therapy, the importance of showing patients a picture, the simplicity of report interpretation, and patients’ reluctance to do pre- and post-meal tests and log insulin. When all three patients initially began using the meter, they infrequently did paired testing (pre- and post-meal) and rarely logged insulin doses. The reports clearly revealed this and prompted a discussion with patients.
    • (1) In the first case study, Dr. Bode knew within 30 seconds that hypoglycemia was a problem, as average blood glucose minus standard deviation put the patient below 70 mg/dl. He emphasized that the entire process occurred digitally, as the patient uploaded the PDF, Dr. Bode reviewed it, and then called the patient to recommend lowering his basal and increasing his meal-time boluses. Dr. Polonsky chimed in that the graphical reports are really valuable from a psychologist’s perspective as well, and they may even prompt better cooperation and adherence.
    • (2) The problem in the second case study was readily apparent after looking at the modal day report: the patient’s glucose average steadily rose between dinner and bedtime. After Dr. Bode showed the patient the report, he admitted a tendency to overeat at dinnertime. The solution was reducing the glargine dose and increasing the bolus dose at dinner (The patient was resistant to trying GLP-1, though Dr. Bode acknowledged in the session that 5 mcg of exenatide would likely be easily tolerated; Dr. DeFronzo suggested pramlintide might also be useful).
    • (3) The third case was a patient with hyperglycemic values throughout the entire day. The suggestion was to increase basal insulin to get fasting glucose down. Dr. Bode also removed the glipizide and sitagliptin, as the patient had a history of hemochromatosis with cirrhosis and had just had a lacunar stroke.

UNDERSTANDING THE BARRIERS OF PATIENT LOGGING AND PATTERN MANAGEMENT

William Polonsky, PhD, CDE (Behavioral Diabetes Institute, San Diego, CA)

Dr. Polonsky discussed how not enough people with diabetes are keeping complete glucose and insulin logs and the obstacles that are keeping them from doing so. He began the presentation by reminding the audience of the importance of regular SMBG to improve glycemic control. Yet, in one Danish/British study conducted in 2000 (Hansen et al., Diabetes Research and Clinical Practice), nearly a quarter of individuals with type 1 diabetes checked their blood glucose less than once a week and only 39% checked it at least once a day. Individuals with type 2 diabetes had similarly low rates of SMBG in the NHANES study from 1988-1994. Dr. Polonsky also explained that insulin users struggle to keep track of the daily activities that influence their diabetes. Even amongst patients who reported occasional SMBG, only half reported bringing their SMBG data to their medical visits and an equally low percentage actively responded to their SMBG data. Dr. Polonsky believes that the three major obstacles keeping patients from faithfully monitoring blood glucose are: (1) inconvenience; (2) patients do not view it as worthwhile; and (3) patients never go over the data with their HCP. Concluding, Dr. Polonsky asserted that the FreeStyle InsuLinx can help address some of these barriers, especially by making it easier to log data, see trends, and make therapeutic changes.

  • People with type 1 and type 2 diabetes test too infrequently and struggle to log their test results and insulin doses. Dr. Polonsky highlighted a 2011 self-reported survey of 500 people with type 1 diabetes and 504 people with type 2 diabetes. The study found that a large portion of insulin users are not logging their test results and even fewer are recording their insulin dosing. Importantly, the fact that the data was self-reported mean that the percentage of patients who do not log results is probably higher than the table below reflects.
 

Insulin users log their test results using (%)

Insulin users log their insulin doses using (%)

Do not log results

36

51

Electronic logbook

18

17

Paper logbook

46

32

  • Few patients bring their SMBG data in when they meet with their HCP – even when they do, few are actively responding to their blood glucose numbers. In a study of 869 individuals with type 2 diabetes who reported having at least occasional SMBG use, only 50% of noninsulin users (NIU) and 58% of insulin users (IU) regularly brought their SMBG data to their medical visits. Furthermore, these individuals did not take action in response to highs and lows: 56% of NIU and 54% of IU reported that they were just observing their numbers. Even when patients do bring their information in, Dr. Polonsky noted that the data is often sparse, disorganized, and hard to interpret. In response to this data, Dr. Bode asked the audience, “How many of you are still willing to meet with a patient if they do not bring their meter or a log?” Approximately a fifth of the audience sheepishly raised their hands – we would bet the number is actually far higher. We also note that this would be easier for patients if they could see their HCP download it for them, which doesn’t happen in many centers (there are some exceptions to this – we also see this changing through a Helmsley Charitable Trust initiative that you can see more about at http://www.closeconcerns.com/knowledgebase/r/1e0d0724.) Dr. Bode said that if a patient does not bring their meter or a log to an appointment, he sends them home to get it. Dr. DeFronzo jokingly added that Dr. Bode’s office still charges the patient for their visit to ensure that they bring their meter or log the next time.
  • Dr. Polonsky stated that the three major obstacles preventing patients from regularly monitoring their blood glucose levels are: (1) inconvenience; (2) a perception that it is not worth the effort; and (3) their HCP either never looks at or discusses the data. Dr. Polonsky said, “The daily ‘job’ of diabetes is already demanding enough” for individuals before needing to spend the time to regularly check and record one’s blood glucose levels. When nothing is done with this information, patients become further frustrated and discouraged. He quoted one of his patients as saying, “When my results are too high, I just get so mad and disappointed with myself.” Dr. Polonsky repeatedly emphasized that the emotional valance patients attach to their blood sugar numbers causes them to avoid testing their blood glucose and/or act dishonestly about their numbers. To try and counter this, Dr. Polonsky gives his patients stickers to place on their meters, which say ‘Remember, it is just a number.’ Many HCPs also find meters are cumbersome to download and the data patients do record is often difficult to interpret. The result is data that is not reviewed at all, which leadspatients to conclude that the data is not worth gathering (or that it can be gathered dishonestly). In one particularly unsettling case, Dr. Bode mentioned he had a young patient who used her lunch money to pay students at her school to prick their fingers. While she had perfect blood glucose numbers on the meter, her A1c was very high. Dr. DeFronzo concurred and said that he believes many patients want to impress their HCPs more than they want to take care of themselves.
  • Dr. Polonsky believes that patients will check and log their blood glucose more faithfully if it is more convenient to do so and if their HCPs take the time to review the data. Dr. Polonsky’s STeP study found that if physicians have the time to sit with patients and review their data carefully, then patients were more likely to track their blood glucose and insulin dose. A1c levels decreased significantly relative to the control (Polonsky et al., Diabetes Technology & Therapeutics 2011). Dr. Polonsky emphasized that HCPs should use this data not just to answer their own questions, but questions that are important to the patient. Key in all of this is having data presented in a way that highlights trends – such representation would help HCPs and also increase the number of patient ‘ah-ha’ moments. Dr. Polonsky reviewed an example to illustrate this point. One of his patients had a similar meal each morning and measured her blood glucose right before the meal and two hours later for a week. She logged the data in a structured table, making it easy for her to determine her average blood glucose change. She was motivated by the obvious impact her data collection was having on modifying (or in this case maintaining) her treatment and wanted to know what questions they were going to look at next. Dr. Polonsky concluded by saying that he likes that the new FreeStyle InsuLinx system allows data to be collected in a way that causes trends to ‘pop’, and that it would have made his patient’s breakfast experiment even easier.

PANEL DISCUSSION

Panelists: Ralph DeFronzo, MD (University of Texas Health Science Center, San Antonio, TX); Bruce Bode, MD (Atlanta Diabetes Associates, Atlanta, GA); William Polonsky, PhD, CDE (Behavioral Diabetes Institute, San Diego, CA)

Dr. DeFronzo: These cases demonstrate how easy it is to use the instrument. It gives a clear picture and helps patients and providers see what’s going on. Now getting to the questions. In the cases, the patient had a low blood sugar at night and a high sugar during the day. Is that a common situation? How do you handle this?

Dr. Bode: There is no question that often people over basalize themselves. They never take enough meal bolus. It’s a knee jerk reaction to the blood glucose running high – up the basal. What that patient needed to do was lower the Levemir and increase the meal dose. It’s a very common problem in both type 1 and type 2s on MDI. It’s especially common in type 1s on a pump. It’s really a knee jerk reaction from the HCP- go up on the basal. Really, you need to cut the basal and go up on the meal dose insulin.

Q: What are your thoughts on Levemir versus glargine?

Dr. Bode: Our hospital just did a weight loss based switch study in which patients went from Humalog to Novolog and Lantus to Levemir, which made no sense. Both insulins work and it does not matter what type of insulin is prescribed. However, there are some special circumstances, such as young individuals with type 1 diabetes, who only need small doses of insulin. Therefore, you need to give them two small doses because the larger the dose you give, the longer it lasts. The smaller the dose, the shorter it lasts. Levemir can be prescribed for twice a day, while many others cannot

Q: How do you see this system integrated with pump therapy?

Dr. Bode: Lots of pumps are integrated with link meters. OneTouch has the Ping into the Animas system. Medtronic initially had BD, and then Novo Max came in. There was the J&J LifeScan meter and now the Bayer Contour Link. Then you have the FreeStyle meter in the OmniPod system. The major benefit of the FreeStyle InsuLinx system is the paired testing. I’d love to have it link up, but it’s really the ability to get the paired testing. And you don’t need anything – no software. The cable comes with the meter. You can put it on your iPad. You can put it on your Mac device or any PC device. I encourage you to talk to your Abbott rep and go to the booth. It’s amazing and an impressive thing that the software is built in.

Q: I have a patient who refuses to check their blood glucose more than once a day. How would you approach this patient?

Dr. Polonsky: Being a psychologist I would start by asking them questions. I would ask them ‘what is in your way from checking more often?’ Often you will find that your patient says that they do not see the value in monitoring their blood glucose. We need to be able to sell the value. We have to tell them ‘here is how I am going to use that data to make your quality of life better.’ In response to the earlier question I am concerned about how easy it is to overwhelm people. When I think about all the software on this device, and all the possible summary data and graphs, I am concerned that it may overwhelm either the HCP or patients, and we need to be careful about that.

Dr. Bode: You may just want one Snapshot report. Typically you just use a few reports. Meal average, Snapshot, and Logbook. I probably would get rid of the rest. But it only takes about three minutes to set it up.

Q: Dr. Bode, you said if your patients showed up without a meter, you said you sent them home. Can you expand upon this?

Dr. Bode: We mandate that you have a downloadable meter. If you bring in a log that’s perfect, we’ll certainly look at it. But we want the meter to validate it. If a patient doesn’t bring it, our medical assistant says ‘Go home and get it. We’ll fit you in the afternoon.” I cannot help them if they don’t bring the meter in. What am I going to do, socialize with them? If a patient’s A1c was 10%, I might say, ‘Well what were you overnight?’ The patient response might be ‘Well I don’t know.’ And I would say, ‘What were you during the day?’ And the patient might respond, ‘During the day, I’m doing well.’ What do I do in that case? I’ll put them on professional CGM if I need more data. I do that all the time. A picture is worth 1,000 words. Once you see it, it’s a game changer.

Dr. Polonsky: As the psychologist on the panel, I get queasy at the idea of sending patients home if they don’t bring in their blood glucose data. But it does send a message that it’s so valuable. In qualitative studies, we ask patients why they don’t check their blood glucose. The response is ‘No one looks at the data.’ It’s so powerful when patients know that this is so core to what we’re doing that I cannot talk to you without it. That sends a powerful message.

Dr. DeFronzo: You showed it in the STeP study. When doctors looked at the data, the improvement was bigger. Bruce is saying the same thing. That offers a big whack on the side of head when the patient shows up.

Dr. Polonsky: Those patients in the intervention group that didn’t bring in data didn’t show a benefit. That conversation between the patient and provider was critical.

Q: I believe that if the patient is not writing in a logbook, he is unlikely to log data in a meter either. What do you think?

Dr. DeFronzo: There is a big difference between having to write something down and just having to touch a button.

Dr. Bode: I agree with Dr. DeFronzo – with technology you can just touch a button and it will log the data for you; it is seamless. There is a big difference between writing something on paper and having technology record it for you. For example, when I write down notes at these meetings and then get home and look at them I go ‘what are all these notes?’ But if I was using a technological system like this then it keeps all those notes straight.

Dr. Polonsky: We have seen that making data logging easier and more convenient causes a big improvement in the rate at which patients record data.

Q: How often do you get complete data on insulin users?

Dr. Bode: The problem is when people have multiple meters that don’t talk to each other. We’d like patients to have one meter. We also encourage meter companies to talk to each other. People might have four meters, and if you could upload and get them to one database, that’s beneficial. I encourage patients to use one meter. You don’t want them using two or three FreeStyle InsuLinx meters. People always need to carry the same meter and work on that. Whatever makes it easy and simple for a patient, the compliance increases dramatically.

Dr. Polonsky: I almost never see patients with comprehensive data. The reason is that I typically see patients who are struggling. But I love hearing about patients that are doing well.

Q: Earlier in the talk, Dr. DeFronzo mentioned splitting the insulin dose when you get above 60-70 units. What do the other speakers think of this?

Dr. Bode: When you’re taking a dose of insulin and you get above 40 to 50 units of these long acting units, then you do not need to split it. However, when you get to over 100 units, then you clearly have to split it because it will require two syringes. In that case, I will definitely look at and probably modify what they are taking.

Dr. DeFronzo: Both glargine and detemir work for 24 hours. Did you see a tapering of the dose in the last six hours?

Dr. Bode: There is no question that there is dose tapering. It is prolonged a little more with Levemir than glargine. At about 0.1 units per hour, they last six to seven hours. Doses of 0.4-0.8 units per hour last much longer. For type 1s on a low dose, you have to split it a lot.

Dr. DeFronzo: For type 2s, you may be giving what you think is a reasonable dose of insulin, but they’re severely insulin resistant.

Q: Currently, many of the treatment guidelines do not use blood glucose values for adjusting patients’ dose; they focus on A1c. Since we are at the ADA, I was wondering if you would push the ADA to add blood glucose values to the guidelines?

Dr. Bode: Many studies that helped determine these guidelines, including ACCORD, were A1c driven. Yet, one A1c value in one individual may mean something very different in another individual. You have to look at blood glucose levels when making decisions about treatment. There is no question that you need glucose data.

Dr. Polonsky: Another reason to encourage blood glucose monitoring is that it is one of our least used motivational tools. Usually it is a de-motivational tool. But I think what we saw here is that it is valuable for physicians and valuable and impactful for patients when done well.

Dr. DeFronzo: I particularly like that you can look at blood glucose levels over the week because it is easy to see for both the HCP and patient what needs to be done.

Dr. Bode: I agree. When showing this data, it’s key to make sure that you are not blaming the patient or making them feel bad about the data.

Q: We use a dosing tool with an insulin adjusting formula. Will the reports ever be available as a .CSV file that can be turned around and sent back?

Dr. Bode: That’s a good point. That file would go into an Excel file or a database. We do this in our hospital with Glucommander. Actually, we recently got approval for outpatient use of Glucommander for subcutaneous insulin. We’re encouraging meter companies to do provide reports in this format. It’s key to trying to close the loop for patients. We think you need to touch base with people weekly and encourage positive reinforcement. I like the idea uploading from home and sending the PDF – I love the digital format and using an automated system for dosing. Most people just don’t know how to adjust insulin appropriately.

Corporate Symposium: Setting the Stage for Better Type 1 Diabetes Care - Live Clinician-Patient Conversations on the Challenges of Continuous Insulin Infusion (Sponsored by Lilly USA)

FACULTY PRESENTATIONS

Steven Edelman, MD (University of California, San Diego School of Medicine, San Diego, CA); Peter Chase, MD (Barbara Davis Center, Denver, CO); Howard Wolpert (Joslin Diabetes Center, Boston MA); Howard Wolpert (Joslin Diabetes Center, Boston MA)

Dr. Edelman gave a comprehensive overview of CSII, beginning with the history of insulin pumps and the main features of current technologies. He noted that pumps have a number of advantages over MDI, including enabling better glycemic control, less weight gain, and more lifestyle flexibility. However, pump therapy also lead to irritation at the infusion site, unexplained hyperglycemia, or improper medication due to technology malfunction or improper use. Dr. Chase discussed who should use insulin pumps, as well as contraindications for insulin pump therapy. He noted that those starting on pumps should be willing to self-monitor blood glucose, have motivation/desire to use a pump, have family support, and be knowledgeable about diabetes. According to Dr. Chase, contraindications to insulin pump therapy include lack of motivation, unwillingness to self-monitor, and non-adherence to injection regimens. Dr. Chase also touched upon suggestions for pump therapy in pediatric populations (including those under six years of age) and the benefits of pump therapy for preventing post-exercise hypoglycemia. Finally, Dr. Wolpert discussed the adoption and use of pump therapy and emphasized that physicians need to focus on engaging patients in their own self-care. He opened by noting that while diabetes technology can make treatment more manageable, it also imposes additional burdens that physicians should consider when individualizing treatment approaches. Dr. Wolpert cautioned that “early young adults,” people aged 18-25 years, balance competing priorities that can distract them from diabetes care and may not be strong candidates for CSII use. Like Dr. Chase, he underscored that patients who adopt pump therapy should be skilled at diabetes self-management. Dr. Wolpert concluded his talk with a brief discussion on the role of CGM in diabetes management, noting that while CGM use has several advantages, patients must have realistic expectations of the technology and an understanding of its drawbacks.

 

LIVE CLINICIAN-PATIENT CONVERSATIONS AND DISCUSSIONS

Dr. Chase’s patient, 17-year-old Monica, discussed her twelve-year experience with insulin pump therapy and the situations she expects to face upon entering college this fall. Monica stated that her favorite aspect of pump therapy was that it allowed her to snack often without worrying about injections. She noted that this advantage would be particularly useful during college because students often eat at irregular times. Dr. Wolpert’s patient, Paul, then touched upon the burdens associated with use of CGM and pump use, how he uses the technologies in conjunction with exercise, and how he uses the data from his CGM. Finally, Dr. Edelman’s patient 76-year old patient Barbara (who was misdiagnosed with type 2 diabetes at 64 by five physicians before being correctly diagnosed with LADA) described how she has adjusted to a CGM and pump, despite not being “tech savvy.” Throughout her comments, she stressed the importance of doctors listening to their patients.

PANEL DISCUSSION

Peter Chase, MD (Barbara Davis Center, Denver, CO); Howard Wolpert (Joslin Diabetes Center, Boston MA); Bruce Bode, MD (Emory University School of Medicine, Atlanta, GA); Steven Edelman, MD (University of California, San Diego School of Medicine, San Diego, CA); and their patients.

Q: How often would you start a patient on pump therapy with saline first?

Dr. Edelman: I typically do not and I don’t think I ever have. Because by the time I think a patient would do well on pumps, I think using saline not needed in my experience.

Dr. Wolpert: It varies. With some patients there’s value in using saline so that the patient can get used to using the pump technology and the controls. But for most patients, we put start them with insulin. But it’s something that should be individualized for each patient.

Monica’s Mom: Monica was five when she started the pump and before we let her use it, my husband and I wore the pump for a few weeks to understand how it works. That process helped us a lot. It was invaluable.

Dr. Bode: I encourage all HCPs to wear a pump and to wear a sensor so you understand what it is like. Especially when starting children on a pump, it’s great to put the parent on the pump first.

Dr. Chase: By our standard, almost everybody does a week or so of saline and performs a set change at home before they come back and start insulin with the pump . If they’re coming from out of town, maybe they’ll start insulin right away, but 98% of our children begin on saline.

Q: How early can you start pump therapy after diagnosis? Would you start a patient within six months of being diagnosed?

Dr. Chase: There are studies that I am involved in that are putting newly diagnosed patients immediately on the pump. The reason is because when someone is recently diagnosed, the liver is producing huge amounts of sugar because there is no insulin. We hypothesize that by turning off the liver early, we might prevent some of the glucotoxicity that destroys the islet cells and we can help people produce insulin for longer periods. If we show that we can shut off the destroying of the islet cells by glucotoxicity, everyone may go to the hospital after being diagnosed for a period of closed-loop pancreas before going home on the pump to keep their pancreas working. If they do that, they may be less likely to have hypoglycemia and develop complications.

Dr. Edelman: when it comes to type 1 diabetes, and maybe all diabetes, there really should be no hard and set rules for anything. I don’t think there should be a set time frame or age limit for pump therapy. I have a patient that has done extremely well – he’s blind and he’s had incredible control over his insulin pumps. I don’t think there should be any limits.

Q: Any data on pump therapy with the elderly?

Barbara: As the patient population matures and there are more of us diagnosed as type 1 later in life, there will be a greater need to include elderly in clinical trials.

Dr. Bode: Dr. Wolpert and I were in a JDRF group that was studying CGM and we only had 5% elderly who were on CGM and so it was hard for us to submit the data to FDA.

Dr. Edelman: I’ve been to the Medicare hearings in Baltimore and something I don’t understand why they think that diabetes is easier to control when you go from 64 to 65. When I went to San Diego as a fellow in 1987, there was a group called insulin pumpers. They are still a strong group and one thing I noticed the last time I was there was that many members are starting to reach retirement. It’s nice to know that people are living long enough to be in that age group, but the government isn’t prepared.

Q: In the previous presentation, one of the contraindications for pump therapy was hypoglycemia unawareness if you live alone. I have a patient had hypoglycemia unawareness who lives alone and I thought pump therapy would be good for her. Am I wrong?

Dr. Edelman: A meta-analysis of over 15 studies showed that the chances of having a severe hypoglycemic episode are lower with a pump. I think that the first step for that patient would be CGM. That would have a more beneficial effect than a pump.

Dr. Bode: I think that’s a mistake in the slide. Many people who live alone wear a pump. I agree that you should use a sensor. Always, when starting therapy, you should decrease the insulin dose just to be safe.

Q: There is no insulin approved for under age 3. What should pediatric endocrinologists do?

Dr. Chase: This has been a real problem in pediatrics. Previously there was no approval of any type 2 agent for pediatrics. So you use these medicines by “physician discretion”. Now the FDA is starting to tell companies that they have to prove that their medications are safe in children and I applaud this approach by the FDA. While it may make the development process longer, it’s a good precaution to take.

Dr. Bode: Insulins have be proven safe through multiple analyses and tests – that’s been done. The problem is that when people are younger than age three, it’s harder to get them to enroll in a trial.

Q: Pumps cost more than injections. The President’s address talked about tsunami of diabetes coming to the US and the rising cost of healthcare. Do you think you’re wasting the government’s money by being on a pump?

Barbara: I’m paying for my pump, so I’m not wasting the government’s money. Monica: No. I think it’s definitely worth it. It helps me out.

Dr. Bode: So basically, it’s like saying, “Do you treat breast cancer in a women?” If you look at cost

effectiveness data, the cost of CGM and pumps are clearly within the guidelines of paying for healthcare. If you have to live day to day with diabetes and try to manage with multiple daily injections, you see that it’s not an easy disease to live with. There’s quality of life involved. That’s one of the goals of diabetes management – trying to live a reasonable quality of life.

Q: Why does giving insulin 10-15 minutes before a meal increases glycemic control while giving it after the meal or just before the meal doesn’t?

Dr. Chase: If you’re really trying to lower A1c – and this is more of an issue in teenagers when hormones are increasing and the risk of complications goes up dramatically –giving a bolus 15 - 20 minutes before a meal helps the glucose levels stay below 180 mg/dl. The ADA said that people should not have glucose levels above 180 and there’s no way you can keep it below 180 without giving a bolus before you eat.

Q: Paul when you were younger, you rebelled and did not adhere to your regimen. Did it help when HCPs told you that if you don’t manage your disease, you’ll get major complications?.

Paul: No, not at all. You don’t internalize it until you actually see something happening right in front of your face. For a lot of people, it takes a sign to motivate behavior change and that’s certainly true for diabetes. It’s the same decision balance model as any other sort of behavioral activity. Are the benefits going to outweigh the effort it takes to accomplish a task? Even now, I do things that I know will raise by blood sugar, but I make that decision.

Q: When will we have closed loop out there?

Dr. Chase: It will come in parts. The first part with be turning off pumps in response to low glucose levels. That should occur in one to two years in the US. The second part will be turning off the pump with predicted low sugar levels, which will likely take three to four years.. The third part is just having closed loop when someone is lying in bed at night. The mean A1c for teenagers in the type 1 diabetes registry during the past year has been 8.6-8.7% - this is where we were 20 years ago. We have not made a lot of progress. I think that if you can even control that for the hours that they are asleep, that will help. I think it will come in stages and it will be a couple of decades before we see it fully complete.

Dr. Wolpert. I agree that it will come in steps. The pumps today are essential the same as those 20 years ago, with changes in ease of use and delivery. I think for the next couple of years, we’re talking about proof of concept studies to show that the technology will be effective.

Dr. Bode: I think capitalism is great – there’s a lot of competition out there. People are working to make faster insulin and different types of sensors. Outside the U.S., I think it the closing of the loop will happen a lot quicker. JDRF said it will have a closed loop system somewhere in the world within 4 years. It will probably not be in the U.S. though. Clearly, I think everything will get better and I think its great to have competition out there. Clearly you will see a lot happening in the next few years. Hopefully Barbara won’t have to pay for sensing in the future. We’re trying to show that sensing really saves money. The registry run by the Helmsley shows that when you go on pump or CGM, the A1c is 1 point lower than with MDI, though the patients may be more compliant to begin with.

Private Event: Hospital-Based Glucose Control (Sponsored by Echo Therapeutics)

COMPANY INTRODUCTION

Pat Mooney, MD (CEO and President); Wayne Menzie (Director of Technology and Clinical Development); Rajko Ilic (VP Product Development, Echo Therapeutics, Philadelphia, PA)

Dr. Mooney welcomed roughly 30 attendees to Echo’s symposium, held in a spacious gallery at the Pennsylvania Academy of Fine Arts. Standing beside Thomas Eakins’ famous painting of Dr. Samuel Gross, Wayne Menzie reviewed the clinical need for glucose control in the ICU, as pioneered by Dr. Greet Van den Berghe and still widely endorsed by clinical organizations worldwide (even though targets are no longer as tight as proposed by Dr. Van den Berghe, following the unsuccessful performance of the intensive control arm in NICE-SUGAR). Rajko Ilic then discussed Echo’s product development efforts, which have been informed by multiple interviews and focus groups with nurses and physicians. He closed with a video demonstration of Echo’s Symphony transcutaneous continuous glucose monitor (tCGM) as it would be used in the ICU. After the nurse has mechanically abraded the skin with the Prelude SkinPrep system, the sensor is worn for 24 hours, during which it wirelessly transmits CGM data every minute to a nearby data display screen (which alarms audibly and visually in hyper- and hypoglycemia).

PANEL DISCUSSION

Moderator: Pat Mooney (Echo Therapeutics, Philadelphia, PA)

Anthony Furnary, MD (Starr-Wood Cardiac Group, Portland, OR); Jeffrey Joseph, DO (Thomas Jefferson University, Philadelphia, PA); Stanley Nasraway, Jr., MD (Tufts Surgical Center, Boston, MA); Eileen Donnelly, RN, BSN (Thomas Jefferson University, Philadelphia, PA); Samir Farah (Echo Therapeutics, Philadelphia, PA)

The majority of the symposium consisted of a star-powered panel discussion featuring Dr. Anthony Furnary (the father of tight glucose control in cardiac surgery), Dr. Jeffrey Joseph (a founder of Animas and frequent consultant on glucose monitoring to the FDA), and Dr. Stanley Nasraway (head of Tufts’ surgical ICU, where he has led the center’s adoption of tight glycemic control). The panelists outlined various success factors for inpatient CGM, including: some combination of point and trend accuracy (with the details to be discussed at an FDA public workshop on June 25), invasiveness (an area where Dr. Furnary said Echo is the clear leader), cost, and interactivity (or “fussiness,” a quality with which clinicians cannot abide). Frequent analogies were drawn to pulse oximetry, which was initially used as an adjunct to point estimates of blood oxygenation in the ICU, then became adopted more and more widely in other areas of the hospital as providers perceive its clinical value and payors/administrators recognize its cost-effectiveness. The overwhelming sentiment from the panelists was that they and their colleagues are hungry for inpatient CGM and eager for regulatory approval – as Dr. Joseph noted, “This is the first technology I’ve been involved in where the end user wants the technology and we can’t get it through.”

Questions and Answers

Q: I think in the past few years, widespread consensus has developed that we need CGM in the hospital. A number of CGM sensors are already approved, at least for ambulatory use. There is a poster on Sunday showing the FreeStyle Navigator in the ICU. Multiple products will be commercialized for inpatient use. From your perspective as clinicians, what are the top three factors that would make you decide on which CGM sensor you would use? What do you think will make Symphony attractive in the coming shootout?

Dr. Nasraway: The issues you raise are true not just for CGM but for a wide panoply of devices in ICU patients. Often these are among the most expensive products in the hospital – how do we decide among devices, or whether to get devices at all? The advantage of the Prelude/Symphony system is that it’s completely noninvasive. When I was first asked to help Echo in 2005 the transcutaneous penetration used ultrasound, which was finicky and more expensive. Clinicians found it intimidating, awkward, and unlike other products they use. The new handheld applicator that vibrates like electric toothbrush is much easier to use. It is also a less expensive and more effective abrasive maneuver. To influence us at Tufts, it would have to be combination of low-hassle – nurses will pitch fussy technologies out the window; in the past we have bought products and then not used them because the form factor wasn’t good enough. I think some sort of reasonable acquisition and maintenance cost will also be important. Healthcare is right up top in major issues facing Massachusetts, as it is nationwide. Just today the Massachusetts House of Representatives passed a measure to reduce paybacks to providers, and to assess a 20% luxury tax on hospitals who overcharge. So I think the two main issues are fussiness and cost, with reasonable accuracy.

Ms. Donnelly: I think the most noninvasive, non-cumbersome monitoring device will be key. Patients are already inundated with central lines, and often they are intubated and on ventilators. We don’t want something to make it more cumbersome.

Q: Are those real savings in your mind?

Ms. Donnelly: We now do Accu-Chek tests every two hours, and every one hour for patients on insulin. This is a huge amount of time, and the repeated fingersticks are painful for patients.

Q: What do you think it will take to convince both CMS and insurers to switch from direct blood sampling to the next generation of CGM?

Dr. Furnary: Just cost-effectiveness.

Your first question was the best one – what do you see the shootout looking like. I think it will come down to four factors: 1) invasiveness, 2) accuracy, 3) cost, 4) interactivity. Right now cost and interactivity are taking a back seat. Companies are either heading down accuracy and not worrying about invasiveness, or like Echo – focusing on invasiveness and working to meet the current standard of accuracy. Once approved, each will go for the other element – the highly accurate sensors will move toward less invasiveness, and Echo will improve its accuracy. Only one company will win noninvasively, and that is Echo. With continuing modifications of the algorithm, it will become more and more accurate. Thus I think Echo is the most well positioned to be a competitor. The other elements come down to cost-effectiveness for administrators and ease/interactivity for nurses. In a really good setting, a BG measurement takes five minutes – one twelfth of an hour. We are spending between $170 and $250 per patient per day to monitor blood glucose.

Dr. Nasraway: It’s not just in strips and the monetary price of nursing time – I think the greater gain is that if nurses are not using 1980s-era technology to measure intermittent glucose, they can take the time to be a real nurse at the bedside – this will add unmeasured benefit.

Ms. Donnelly: I think the time issue is critical, too – if you have to make an insulin drip infusion change based on a blood glucose reading that is supposed to be taken at 2 pm, you may not actually have a chance to take the measurement until 2:20 – whereas if you have something that takes a continuous readout, you could act faster.

Dr. Joseph: In our facility we take 32,000 glucose tests per month – 4% of these readings are below 70 mg/dl, and 20% are above 200 mg/dl. CGM trend data will become widely seen as critical. As it was with pulse oximeters in the ICU, once you can see trend data, you’ll never want to go back to ‘flying blindly’ with intermittent measurements.

Q: In the CGM oral presentations, a competitive company called into question the data of other companies [Editor’s note: see coverage of Dr. David Price’s oral presentation above]. Do you clinicians have opinions on the quality of Echo’s data and what will be good enough to get to market?

Dr. Furnary: I have been doing this for 20 years. I think anything as good as point-of-care glucose is good enough – I get point-of-care measurements every one or two or four hours, and we avoid hypoglycemia 99% of time with a target range of 70-110 mg/dl. For me, even the current accuracy is enough with our protocol. This will vary by center… but what really matters is what FDA says. Go to Maryland on June 25th to find out – we still don’t know what benchmark the agency will set. [Editor’s note: for more on the FDA’s upcoming public meeting on clinical study design and performance of hospital glucose sensors, see http://www.fda.gov/MedicalDevices/NewsEvents/WorkshopsConferences/ucm304483.htm.]

Dr. Joseph: Regulatory agencies have recognized that it’s not just about the point accuracy, it’s about the trend accuracy. We have to figure out how to show trend accuracy to get approval. Once a system is approved, clinicians will use it and develop protocols. The FDA is on our side; they would love to see inpatient CGM approved.

Dr. Mooney: You are on the FDA panel on June 25. Dr. Joseph was on a panel at ISICEM in Brussels that seems that it will lead to recommendation that point accuracy be in the 12%- 13% range, on the basis that frequent measurements would reduce the need for point accuracy.

Dr. Furnary: Where do you think FDA will come out on continuous monitoring requirements?

Dr. Joseph: They are not sure where they stand, looking for us to give them advice, hence this workshop. It will be a combination of point and trend accuracy, to be defined.

Artificial Pancreas and Pumps

Oral Sessions: Closed Loop Systems

STATE OF THE ART LECTURE — PITFALLS IN THE DEVELOPMENT OF A CLOSED LOOP

William V. Tamborlane, MD (Yale University, New Haven, CT)

After reviewing the state of closed-loop research (including his own group’s recent work with Insuline’s infusion-site-heating InsuPatch), Dr. Tamborlane proposed that the biggest remaining obstacle is to ensure that a system’s malfunctions will not result in harmful overdelivery of insulin. Underdelivery of insulin has also been raised as a concern, however, most notably in the FDA’s review of Medtronic’s low glucose suspend (LGS) Veo pump/CGM. To investigate whether erroneous insulin suspensions could reasonably cause diabetic ketoacidosis (DKA), Dr. Tamborlane and his colleagues conducted a crossover-design experiment in pump-using adolescents and young adults with type 1 diabetes (n=14), in which participants’ pumps were suspended for two hours on random nights. On the mornings after these suspension events, significant rises were seen in the levels of both blood glucose (199 vs. 151 mg/dl) and the ketone beta-hydroxybutyrate (0.11 vs. 0.07 mmol/l). Encouragingly, however, the mean rise was not clinically significant, and the highest single ketone reading (0.6 mmol/l) corresponded to only mild ketonuria. Dr. Tamborlane said that every sane person in diabetes supports LGS, provided the system is assured to be safe. Based on his own group’s results and the findings from the inpatient ASPIRE study (presented by Dr. Satish Garg later in the session), so far he is taking a positive view on the safety and efficacy of the Veo as the first small step toward the goal of an artificial pancreas (which to him means a system that provides fully automated, 24-hour control as well as remote data transmission).

Questions and Answers

Q: You said that the ultimate obstacle is overinfusion of insulin. Is that really the ultimate obstacle? I would wonder if it’s actually cost, given what is happening with payors for type 1 diabetes.

A: Dr. Hirsch is obviously from a blue state, where the lowest common denominator of care is delivered for everyone. I am really not prepared to discuss cost, though I think all the company people need to pay attention to this. We all want to make things available and hopefully this won’t be an obstacle.

Q: It is gratifying to see that overnight LGS is safe. But I am still concerned that people will make behavior changes if they think they are protected, and perhaps give higher boluses at dinnertime than they otherwise would. Then if the system is reading low, there could be problem.

A: There is a big turnout at this session because we all envision automatic glucose control. This was Medtronic’s smallest step they thought they could get through the FDA – we would hate to get hung up on the Veo. We would like to have suspension for predicted lows, so you don’t even get an alarm – alarm fatigue is a big problem. Another FDA concern with the Veo was that A1c would increase A1c. But even in the JDRF CGM trial, we said we would accept an A1c rise of 0.3% in well-controlled patients if it meant less hypoglycemia. Another consideration is that use of CGM in individual patients is a delicate balance between benefits and hassles (which are the main reason people stop using it now). Increasing the benefits would increase the likelihood that patients would use these technologies. There are lots of downstream consequences, intended and unintended, of closed-loop research.

Q: Might GLP-1 agonists be a way to suppress glucagon in type 1 diabetes?

A: We are very excited. I see friends from JDRF in the audience. We have proposal to titrate GLP-1 receptor agonist dose in both open- and closed-loop. That was a very nice softball question.

AUTOMATIC INSULIN PUMP SUSPENSION FOR INDUCED HYPOGLYCEMIA: THE ASPIRE STUDY (221-OR)

Satish Garg, MD (University of Colorado Denver, Aurora, CO)

Dr. Garg reviewed the in-patient ASPIRE study of Medtronic’s low glucose suspend (LGS) pump/CGM system. As a reminder, the crossover-design study showed that for patients undergoing exercise- induced hypoglycemia, suspending insulin delivery restored normoglycemia faster (and with similar negligible risk of rebound hyperglycemia) compared to when the LGS feature was not activated and insulin continued to be delivered (see additional details in our coverage of Dr. Garg’s ATTD 2012 presentation at http://bit.ly/ybzJCD).

THE ORDER EFFECT OF THE IN-CLINIC ASPIRE STUDY: HYPOGLYCEMIA BEGETS HYPOGLYCEMIA (220-OR)

Satish Garg, MD (University of Colorado Denver, Aurora, CO)

Immediately following his review of top-line ASPIRE data, Dr. Garg explained that patients who underwent LGS-on experiments on their first study day (Group A) recovered from hypoglycemia much faster than those whose LGS-on experiments came on the second study day (Group B). Indeed, the between-group difference was 63.7 minutes (107.8 vs. 171.5 mg/dl; p<0.01). The difference seems to be due to the fact that on the second experimental day, patients were still affected by the hypoglycemia induction from the first experimental day, even following the 3-to-10-day washout period. (The mechanism may have involved depletion of glycogen stores or loss of counterregulatory response – unfortunately data were not collected to investigate either of these hypotheses.) Dr. Garg proposed that future crossover-design studies of hypoglycemia induction should use a longer washout period to avoid order effects. He also hypothesized that real-world use of LGS, by reducing exposure to low blood glucose, could potentially help halt the vicious cycle of hypoglycemia begettting hypoglycemia. However, as Dr. William Tamborlane noted during Q&A, exercise-induced hypoglycemia is quite different from low blood sugar during deep sleep, so the only way to truly understand the latter is through the ongoing outpatient ASPIRE study (the pivotal trial for the US version of the Veo, which recently began the FDA review process; see our coverage of Medtronic’s recent analyst day at http://bit.ly/LhilPw). Dr. Garg said the outpatient results could be ready in time for next year’s ADA – we’re already looking forward!

  • The Medtronic-funded ASPIRE study was designed to assess the low glucose suspend (LGS) feature of the Veo pump/CGM in response to exercise-induced hypoglycemia. The inpatient study’s participants (n~50, mean age 34, mean A1c ~7.9%, BMI 27 kg/m2) first underwent a six-week washout period involving several clinic visits. As for the protocol itself, patients entered the clinic early in the morning, after an overnight fast. (In order for an experiment to begin, patients were required to have blood glucose of 100-140 mg/dl.) Patients then exercised until YSI-measured plasma glucose fell to 85 mg/dl, at which time they stopped exercising and were observed. Once the sensor glucose hit 70 mg/dl, the low glucose suspend feature was activated, suspending insulin delivery for two hours. (If the patient failed to drop to 85 mg/dl within 10-15 minutes of exercise, he or she would take a break before re- initiating exercise up to five more times in the visit.) If a patient’s YSI blood glucose value ever fell below 50 mg/dl or rose above 300 mg/dl, the experiment was immediately stopped and treatment administered – Dr. Garg said that between the various requirements, roughly one-third of the inductions had to be repeated. The other arm of the crossover study design used the same exercise protocol, but without the LGS turned on. Patients were randomized to receive either Group A (LGS-on first) or Group B (LGS-off first), and the visits were separated by a washout period of 3-10 days.
  • Participants who underwent LGS-on experiments on their first in-clinic day (Group A) had significantly shorter durations of hypoglycemia than those whose LGS-on experiments occurred on the second day (Group B) – the between-group differential was a whopping 63.7 minutes (107.8 vs. 171.5 mg/dl; p<0.01). That’s a big drop – and also a good reminder of how prevalent hypoglycemia is. People in group A also required significantly fewer inductions before a successful LGS-on experiment occurred (0.36 vs. 1.57 prior inductions1, respectively; p<0.001). Related to their having fewer prior inductions, those in group A also had less cumulative duration of YSI-measured hypoglycemia prior to a successful LGS-on experiment (16.6 vs. 204.6 minutes; p<0.01). Statistical analysis showed that the between-group difference in hypoglycemia duration could not be attributed to sensor glucose rates of change, exercise duration, or spontaneous hypoglycemia (as measured by the area of the curve spent with glucose below 70 mg/dl in the two days prior to successful induction): the p-value was below 0.3 for all three. 
  • Dr. Garg presented a graphic of every-30-minute mean YSI values to illustrate that LGS-On gave a better trajectory than LGS-Off no matter which days the experiments were conducted; however, other measures seemed to make the order effect appear more severe. At ATTD 2012, Dr. Thomas Danne showed that duration of hypoglycemia was not statistically different between the LGS-On and LGS-Off experiments run on day two (171.5 min for Day 2 LGS-On vs. 167.7 min for Day 2 LGS-Off min; p=0.4)). For more details from Dr. Danne’s talk, see page 5 of our coverage at http://www.closeconcerns.com/knowledgebase/r/aa724a9a.

Questions and Answers

Q: You hypothesize that the order effect is due to impaired counter-regulation. That raises the question of efficacy in those already known to have impaired counter-regulation. In deep sleep, we have shown – and Dr. Cryer has confirmed – that you don’t get a counterregulatory response, even if you otherwise would. This finding does support the need for an at-home experiment to see what happens when people are sleeping and not on a treadmill, which is totally different.

A: We hope maybe by next year’s ADA to have results from the in-home study.

Q: Your data captured only events where YSI crossed below 85 mg/dl. I guess this means that events where the sensor might have erroneously suspended at a higher glucose level are thus not captured. Can you explain this rationale?

A: I wish I could go into the details of the regulatory hurdles. The protocol was written so strictly that observation could start only if YSI glucose was below the given number. However, it is likely that many times the sensor glucose would caused LGS events earlier or later. In hindsight we should have had more leverage.

Q: So for the LGS-Off condition, how was basal insulin delivery handled?

A: It was continued, even when the sensor glucose fell below 70 mg/dl. There was no choice in that.

AUTOMATED MANAGEMENT OF BLOOD GLUCOSE IN CHILDREN WITH TYPE 1 DIABETES USING A BI-HORMONAL BIONIC PANCREAS (222-OR)

Edward R. Damiano, MD (Boston University, Boston, MA)

Dr. Damiano gave us an “anecdotal” perspective on his team’s research in this thrilling talk, which included a first-ever look at the iPhone-driven system that will be used in upcoming five-day closed-loop experiments. His group has developed a universal algorithm that can be initialized with only one variable (body weight) and thereafter is robustly adaptive to account for the wide inter- and intra- individual variability in insulin needs (two-to-three-fold changes in insulin dosage are possible within 12-to-24 hours!). Pre-meal priming boluses (which require the patient simply to input whether they are about to eat a low-, mid-, or high-carb meal) allow more robust adaptability, but systems without pre- meal priming offer performance as well. Dr. Damiano looks forward to trying out the group’s latest system in a five-day transitional study as early as late fall/early winter 2012, pending FDA approval of the device (the IDE will be submitted in six-to-eight weeks). He piqued the audience’s interest still further by showing off the actual physical system that will be used in the study, which he said was turned on for the first time earlier that morning (Dr. Damiano was actually wearing a Navigator CGM during the presentation and showed the audience his streamed, real-time glucose information). The hardware consists of: two Tandem t:slim insulin pumps delivering insulin and glucagon, an iPhone to run the controller algorithm and communicate with the pumps via low-energy Bluetooth (no laptop or tablet required), and a Navigator CGM. (The researchers will also run experiments with a platform using a Dexcom G4 CGM, based on their findings that the G4 has superior day-two accuracy vs. the Navigator – see Dr. Steven Russell’s presentation from ADA 2012 Day #1 at http://bit.ly/LMjC1B). We’ve always been impressed with Dr. Damiano’s data-driven, nose-to-the-grindstone approach and hope FDA is receptive to the new device and ambitious study.

COMPARISON OF TWO CLOSED LOOP ALGORITHMS WITH OPEN LOOP CONTROL IN TYPE 1 DIABETES (224-OR)

J. Hans DeVries, MD (Academic Medical Center, Amsterdam, Netherlands)

Building on Dr. Eric Renard’s preliminary presentation of findings at ATTD 2012 (see page 19 of our coverage at bit.ly/ybzJCD), Dr. DeVries presented the intent-to-treat analysis of the CAT Trial – a crossover-design comparison of 24 hours under open-loop control, closed-loop control with the Padova/Pavia/UVa (iAP Consortium)’s MPC algorithm, and closed-loop control with the Cambridge University team’s MPC algorithm. Both algorithms performed similarly and were not statistically significantly different from open loop therapy with respect to the study’s primary endpoint, time in range (70-144 mg/dl in fasting periods; 70-180 mg/dl in postprandial): 58% (Cambridge), 59% (iAP), and 62% (open loop). The good news was that both MPC algorithms offered significant improvements for hypoglycemia <70 mg/dl: 2% vs. 2% vs. 6%. Less encouragingly, the MPC algorithms also caused statistically significantly higher mean glucose (149 vs. 148 vs. 126 mg/dl) and significantly more time spent in hyperglycemia. However, Dr. DeVries noted that recent advances (e.g., more accurate CGM sensors) could allow researchers to make their algorithms more aggressive without compromising safety.

  •  The    CAT Trial compared closed-loop control using MPC algorithms from Cambridge and Padova/Pavia/Montpellier (iAP). Participants with type 1 diabetes used the Insulet OmniPod, a Dexcom Seven Plus, and either the UCSB artificial pancreas system or manual validation from a nurse. Patients were studied in three non-consecutive 24-hour periods under three different conditions (open-loop, closed-loop with Cambridge algorithm, closed-loop with iAP algorithm; order of the conditions was randomized). Six centers took part in the study, with eight patients per center participating, and a total of 142 experiments were performed.
  • The study included meal, rest, and exercise periods with time in range as a primary endpoint. Participants entered in the evening, ate dinner, spent the night, had breakfast and lunch the following day, and concluded the final afternoon with an individualized exercise session. Target range was defined as 3.9-10 mmol/l (70-180 mg/dl) in the first three hours after a meal and 3.9-8 mmol/l (70-144 mg/dl) otherwise. In this presentation Dr. DeVries discussed only intent-to-treat data; per-protocol results (excluding data contaminated by errors from the CGM, pump, or study personnel) will be shown at EASD 2012.
  • Compared to open-loop control, both the Cambridge and iAP algorithms conferred significantly less time in hypoglycemia below 70 mg/dl, though this came at the expense of statistically significant increases in mean glucose and time spent in hypoglycemia. A trend toward significance was observed in time spent at or below 50 mg/dl, and no significant change was observed in the primary endpoint of time in zone. Dr. DeVries noted that these results made sense in light of the higher insulin doses used with open-loop control. (The iAP and Cambridge algorithms were not statistically significantly different with regard to any of the glycemic measurements analyzed, though more insulin was dosed with the iAP algorithm (1.7 vs. 1.6 U/hour; p<0.05).
 

Open Loop

iAP

Cambridge

Overall P- value

% Time in Range

62.6

59.2

58.3

0.377

% Time < 70 mg/dl

6.4

2.1

2.o

0.001

% Time ≤ 50 mg/dl

0.92

0.43

0.17

0.072

Mean Glucose (mg/dl)

 

126

 

148

 

149

 

0.001

Mean Insulin dose (U/hr)

 

1.8

 

1.7

 

1.6

 

0.001

  • The CAT Trial suggested an inverse relationship between hypoglycemia and mean glucose, but Dr. DeVries said that recent technological advances might enable closed-loop systems to improve both hypoglycemia and mean glucose simultaneously. He explained that in CAT, both the Cambridge and iAP algorithms had been “detuned” as a safety measure to account for the effects of exercise and the “suboptimal” function of the Dexcom Seven Plus. He said that fortunately, the greater accuracy of newer CGMs (e.g., the Dexcom Gen 4) could enable the algorithms to be made more aggressive in reducing hyperglycemia while still also preventing hypoglycemia. As Dr. Hirsch noted during Q&A, data from the Helmsley Charitable Trust suggests that hypoglycemia is similarly prevalent across the type 1 diabetes population regardless of A1c. Thus, a great deal of patients could benefit from a system that simultaneously improves both mean glucose and incidence of lows.

Questions and Answers

Q: In the data from the T1D Exchange, we actually don’t see an increased prevalence of hypoglycemia at lower A1c – the rate of hypoglycemia is similar across all A1c ranges. If we assume that this is a true observation, would closed-loop systems maybe be better for those specifically with higher A1c?

A: I think the algorithms could be adapted to more-accurate CGM systems in a way that makes it possible to lower mean glucose with these algorithms. Such a possibility has been suggested in earlier research as well as some that is about to be published.

FACTORS AFFECTING PERFORMANCE OF OVERNIGHT CLOSED-LOOP INSULIN DELIVERY IN TYPE 1 DIABETES (T1D) (219-OR)

Marianna Nodale (University of Cambridge, UK)

Ms. Nodale presented an interesting retrospective review of closed loop studies from Cambridge. The study looked for correlations between overnight closed loop performance and various factors: age, A1c, total daily dose, insulin absorption rate (time to peak of plasma insulin; Tmax), controller effort (the ratio of total insulin delivered during closed loop to the basal profile pre programmed on a patient’s pump). Controller effort was the factor most correlated with overnight closed-loop performance – more controller effort was associated with a higher mean plasma glucose, a higher standard deviation, and less time in target range (70-145 mg/dl). Ms. Nodale explained that the Cambridge controller algorithm has a safety feature that is based on the basal profile of patient – when controller effort was high, the safety feature prevented aggressive insulin dosing to bring down hyperglycemia. Aside from controller effort, better time in target was associated with older patients and those with high total daily doses, while more time spent in hypoglycemia (<70 mg/dl) was associated with a higher A1c and a higher Tmax. Ms. Nodale also emphasized the wide interindividual variability in controller effort (25-200%) and Tmax values – this is now a clear theme we’ve heard in closed-loop talks and we’re glad to see researchers better characterizing and hopefully dealing with this challenge.

  • Overnight closed-loop performance from seven closed-loop studies (1,612 hours) in 79 patients was analyzed. All studies occurred in a clinical research setting and both manual and automated closed loop delivery were included. An adaptive MPC algorithm was used. Factors considered included age, A1c, total daily dose, insulin absorption rate (time to peak of plasma insulin), and controller effort (the ratio of total insulin delivered during closed loop to the basal profile pre programmed on patient’s pump). Overnight closed loop performance was defined as midnight to 8am.
  • Combined, the seven studies’ participants included 11 children, 44 adolescents, and 24 adults; median controller effort was 110% (i.e., the controller delivered 10% more insulin relative to the pump’s pre-programmed basal rate). Mean age was 20 years, mean A1c was 8%, mean BMI 22.5 kg/m2, mean duration of diabetes was 10 years, mean duration on a pump was two years, and mean total daily dose was 49 units per day.
  • Controller effort was significantly positively related to mean plasma glucose and standard deviation of glucose and significantly negatively related to time in target range (70-145 mg/dl), time below 70 mg/dl, and low blood glucose index (LBGI; a measure of severity and frequency of hypoglycemia). Generally, the controller algorithm did not need to give much more insulin than patients had pre-programmed, but Dr. Nodale explained that there was wide inter-individual variability. In some cases, the controller delivered only 25% as much of the preprogrammed basal profile, while other patients needed almost twice as much insulin. She explained that the controller algorithm has a safety feature that is based on the basal profile of patient. In cases of high controller effort, the safety feature likely prevented the controller from being more aggressive and bringing down hyperglycemia.
  • In addition to controller effort, age, A1c, total daily dose, and Tmax were significantly associated with closed loop performance statistics. Age and total daily dose were significantly associated with time in target range. A1c was significantly associated with time <70 mg/dl, and Tmax was significantly associated with time <70 mg/dl and LBGI.

Spearman’s rho correlations

 

Mean Plasma Glucose

SD of Plasma Glucose

Time <70 mg/dl

Time in target range

LBGI

Age

 -0.03

-0.01

-0.06

0.23*

0.04

A1c

-0.05

0.16

0.24*

-0.19

0.17

Total Daily Dose

 

-0.16

 

-0.15

 

-0.05

 

0.29*

 

0.07

Controller Effort

 

0.51**

 

0.24*

 

-0.28*

 

-0.40**

 

-0.42**

Tmax -0.2 0.06 0.24* 0.05 0.28*

*p<0.05, **p<0.01

Questions and Answers

Q: You pointed out that increased controller effort was associated with higher glucose concentrations. Does that mean that your control algorithm was set quite conservatively?

A: Yes, there is a tradeoff between aggressiveness and risk. The controller is conservatively tuned. In patients who are resistant to insulin or have their basals underestimated, the safety aspect makes the controller too conservative.

Q: Has the algorithm learned iteratively? More recently, is it looking better?

A: Each patient was only studied on one occasion. We had few on more than one night, so we cannot assess how the algorithm will adapt to changing insulin requirements. We suspect it will do better. We will have opportunity to study patients for several days.

Q: Is it valid to compare insulin Tmax between patients who received varying levels of insulin from the controller?

A: Please refer to the poster for how we calculate Tmax. I think so.

DUAL-HORMONE CLOSED-LOOP (CL) SYSTEM IN ADULTS WITH TYPE 1 DIABETES (T1D): RANDOMIZED CROSSOVER TRIAL (223-OR)

Ahmad Haidar, MSc (Ecole Polytechnique de Montreal, Quebec, Canada)    

Mr. Haidar presented results from a 15-patient bi-hormonal overnight closed-loop study from Canada. (We cannot recall having ever seen a presentation from this closed-loop team). He mentioned that this was the first randomized, crossover design comparing open loop control to bi-hormonal closed loop. Participants undergoing closed loop spent 71% of time in target range (72-180 mg/dl), compared to 57% of the time during open loop therapy (p=0.003). Time spent <72 mg/dl declined significantly from 10% during open loop to 0% during closed loop (p=0.01). This is encouraging data and we look forward to hearing more about this team’s work. Due to large differences in study design and methods (e.g., meal size, type, schedule, sensor calibration, algorithm interaction), it is hard to compare this trial to the bi- hormonal work from Dr. Ed Damiano and colleagues in Boston. However, we’re glad to see so much interest in dual-hormone control and look forward to even better performance, a stabilized glucagon in solution, and dual-chamber pumps down the road.

  • This 15-hour randomized, crossover study included 15 adults with type 1 diabetes (9 female, mean age 47 years, mean A1c 8%, mean BMI 26 kg/m2). Patients were admitted to the clinical research center twice and received either closed loop or open loop treatment. The intervention started at 4 pm, 30 minutes of exercise (60% VO2 max) occurred at 5:50pm, a dinner was eaten at 7:20 pm (80g carbs for males and 60g carbs for females), and then a 15g carb snack was eaten at 10 pm followed by sleep until 7 am the next morning. Meals were announced to the algorithm.
  • The algorithm’s input was sensor readings from a Medtronic Sof-Sensor and subcutaneous insulin and glucagon doses were recommended every ten minutes. The algorithm was based on a fuzzy-supervised model-based predictive controller combined with an extended Kalman filter and a set of heuristic rules. Mr. Haidar did not describe how portable or software integrated the system was. Plasma glucose was measured using YSI 2300 STAT Plus Analyzer. Study outcomes are based on plasma glucose readings.
  • Time in target range (72-180 mg/dl) improved from 57% during open-loop therapy to 71% during closed-loop therapy (p=0.003). Time spent under 72 mg/dl declined from 10% to 0% during closed loop (p=0.01), and time spent under 60 mg/dl dropped from 2.8% during open loop to 0% during closed loop (p=0.006). Time spent above target was not significantly different between the two groups. Standard deviation of glucose was 60% lower during closed loop visit (20 mg/dl vs. 34 mg/dl). Insulin delivery and concentration were not significantly different between the two groups.
  • The number of patients that had at least one hypoglycemic event (<54 mg/dl) dropped from 53% during open loop to 7% during closed loop (p=0.02). Nocturnal events declined from 32% to 0% (p=0.07). Exercise induced hypoglycemia was not significantly different between the groups.

Questions and Answers

Q: Thank you for this very important and very impressive study. In some cases, we noticed in our studies that giving small doses of glucagon, resulted in no reaction. Glucose continued to fall. Did this happen in your study?

A: I don’t remember any times where that happened. There was one case of hypoglycemia. The sensor algorithm was not aggressive, so we didn’t have cases of hyperinsulinemia.

Q: Given the instability of glucagon, how did you prepare it in this study?

A: We prepared it at the beginning of the study. It was in the pump for 16 hours.

Q: Since we’re learning more about dual hormones, can we see glucagon resistance develop?

Dr. Ward: When insulin levels are high, the response to glucagon is less. The key is to keep insulin dosing to a minimum. We’re also looking to see if repeated doses of glucagon deplete hepatic glycogen. We’ll be looking at this in a study.

A: Glucagon is a safety layer. With exercise, we had large incidence of hypoglycemia. We had to give glucagon to prevent it.

Oral Sessions: Late-Breaking Abstracts

CLOSED-LOOP INSULIN THERAPY IMPROVES NOCTURNAL GLYCEMIC CONTROL IN CHILDREN <7 YEARS (153-LB)

Andrew Dauber, MD, MMSc (Boston Children’s Hospital, Boston, MA)

Dr. Dauber presented results of an overnight (10 pm – noon), crossover-design closed-loop study in children under seven years old (n=10; mean A1c 8.1%, age 5 years old, diabetes duration 2.1 years; C- peptide levels undetectable). Compared to open-loop therapy, closed-loop control with a FreeStyle Navigator and a PID algorithm led to statistically significantly shorter duration of overnight hyperglycemia > 300 mg/dl (0.18 vs. 1.3 hours) without an increased risk of hypoglycemia. The post- breakfast spike was sharper after breakfast since no pre-meal priming boluses were used, but closed- loop control also improved mean pre-lunch blood glucose values (190 vs. 270 mg/dl). Dr. Dauber believes that the young pediatric population is well-suited to closed-loop control (due to their predilection for hypoglycemia and their unpredictable eating and activity habits) but is underrepresented in current artificial pancreas research.

  • The crossover-design study required study participants (n=10) to spend two consecutive nights in the hospital wearing an Animas Ping pump and two Abbott Navigator sensors, on each leg. (The second sensor was included in the study only as a failsafe and turned out not ever to be necessary – in every experiment, the system was run based solely on input from the sensor worn on the patient’s right leg). During the day between experimental nights (noon to 10 pm), patients were able to use open-loop control, disconnect themselves from the intravascular line used to test blood glucose over night, and walk around the hospital.
  • In the closed-loop condition, insulin dosage was determined based on physician entry of sensor glucose values into an algorithm that targeted 150 mg/dl from 10 pm to 6 am and 120 mg/dl from 6 am to noon. Over night, the algorithm adjusted basal insulin rates every 20 minutes. In the morning, the system delivered microboluses as often as every minute (the system was fully reactive: no pre-meal priming boluses were given). During Q&A, Dr. Dauber said that future versions of the system might incorporate every-minute dosing at night, hopefully enabling lower rates of hypoglycemia at the same mean glucose target.
  • Overnight time in hyperglycemia > 300 mg/dl was statistically significantly reduced with closed-loop control (0.18 vs. 1.3 hours), and overnight time in the target range of 110-200 mg/dl was non-significantly greater (5.3 vs. 3.2 hours). Exposure to overnight hyperglycemia (area under the curve above 200 mg/dl) was also significantly reduced with closed-loop control. Because no pre-meal priming boluses were given, the post-breakfast glucose rise occurred faster among closed-loop patients. However, both groups reached similar peak postprandial glucose values (367 vs. 353 mg/dl), and the closed-loop patients’ glucose fell back toward normal much faster –mean glucose values at noon were 270 mg/dl and 190 mg/dl, respectively. The groups were comparable in instances of hypoglycemia < 70 mg/dl during the observational period (four vs. five events; treated with juice), though five instances of hypoglycemia also occurred in the open-loop rest period between experimental nights (potentially related to the relatively higher rates of insulin given under the closed-loop than open-loop condition, as suggested during Q&A).

Questions and Answers

Q: Did you use the data from both sensors?

A: We used only one – the second was in case the children ripped off one of them. By convention we always used the sensor on the right leg.

Q: You showed a lot of hyperglycemia after breakfast.

A: There is no doubt that pre-meal priming boluses will improve closed-loop therapy. But there are reasons to be concerned about this. It is extremely difficult to predict how much children will eat.

Q: Do you think you should raise the overnight target even higher to avoid hypoglycemia?

A: The 150 mg/dl overnight target was already right in the middle of the ADA’s recommended range. Based on the information from this study, we will refine the parameters for our closed-loop algorithm. I think we will see improvement by switching to once-a-minute adjustments overnight, instead of every 20 minutes. My aim would not be to raise the target but rather to improve insulin delivery.

Q: You wound up giving a lot more insulin in the closed-loop condition. In the open-loop ‘rest’ period, was there a difference in glycemic control after lunch?

A: That was not included in our analysis. Five episodes of hypoglycemia occurred outside the outcome period of the study [i.e., outside of the 10 pm – noon range when the experiments were actually running]. I think a morning snack might help blunt the postprandial hypoglycemia, but we checked blood sugars only at noon, three, and prior to dinner.

Posters

FEASIBILITY STUDY ASSESSING HYPOGLYCEMIA-HYPERGLYCEMIA MINIMIZER (HHM) SYSTEM IN PATIENTS WITH TYPE 1 DIABETES (T1DM) IN A CLINICAL RESEARCH CENTER (CRC) (917-P)

Linda Mackowiak, Daniel Finan, Thomas McCann Jr., Ramakrishna Venugopalan, Howard Zisser, Henry Anhalt

This feasibility study is the first data we’ve seen since the JDRF/Animas partnership formed in 2010 to develop and commercialize a first-generation artificial pancreas device. The Animas system uses a OneTouch Ping pump, a Dexcom Seven Plus CGM, and a hypoglycemia-hyperglycemia minimizer (HHM) algorithm running on a laptop (i.e., controlling to a range of 90-140 mg/dl). The investigational device was tested in the clinic in 13 patients with type 1 diabetes. The 20-hour study included two important challenges to the algorithm: (1) at the breakfast meal, insulin was under-bolused (up to 50%) and (2) at the lunch meal, insulin was over-bolused (up to 50%). The system achieved a mean blood glucose of 165 mg/dl, correlating to a respectable A1c of 7.4%. Time in the range of 70-180 mg/dl was 70% overall, over 80% at night, and less than 1% at values <70 mg/dl. There were no events of DKA or severe hypoglycemia. Although this was just a small feasibility study, we think the data is encouraging and are glad to see the partnership moving forward. While some might say 165 mg/dl is not perfect control, and while we very much agree with this, we believe it’s directionally solid considering the average A1cs in the Helmsley T1D Exchange (over 8%) and the high prevalence of hypoglycemia and severe hypoglycemia (see our ATTD 2012 report at http://bit.ly/zEKWOB). Additionally, we note that participants ate fairly high-carbohydrate meals and the system was thrown curveballs with real world insulin dosing mistakes. Besides fine-tuning the algorithm, we believe the most critical future step is making this Animas system portable – we’ve already seen two cellphone based portable systems that have been tested in trials: Medtronic’s Portable Glucose Control system and the iAP consortium’s Diabetes Assistant (for the basics on both devices, see pages 3-6 of our DTM 2011 report at http://bit.ly/uh3qXd).

  • This study used Animas’ hypoglycemia-hyperglycemia minimizer (HHM) System, which consists of an OneTouch Ping insulin pump and meter, a Dexcom Seven Plus CGM, and the HHM algorithm. The algorithm runs on a laptop and uses the UCSB/Sansum Artificial Pancreas System (APS) platform. The HHM targets a glycemic zone setting of 90-140 mg/dl. The control algorithm automatically adjusts insulin delivery in response to changes in CGM values, as well as predictions of future CGM trends. The algorithm is designed to take action in order to reduce or prevent glucose excursions outside of the target zone.
  • This nonrandomized, uncontrolled, inpatient feasibility study tested Animas’ HHM system over a 20-hour period (clinicaltrials.gov identifier: NCT01401751). Participants’ pump settings were assessed and optimized by investigative staff during the week before the CRC visit. CGM sensor insertion occurred two to three days prior to the study. Participants arrived in the early evening on day one of the study and closed-loop control was initiated at midnight and lasted 20 hours. Breakfast on day two was eaten around 07:00 am and included a manually givenunder-bolus of insulin (up to 50% in some cases). Lunch on day two was eaten around 01:00 pm and included a manually given over-bolus of insulin (up to 50% in some cases). (Rather than just giving the maximum amount of each over- or under-dose, we wish that patient-specific bolus amounts had been given.) Both meals included one gram of carbohydrate (CHO) per kilogram of body weight, up to a maximum of 100 grams of carbohydrates. Regular YSI monitoring was performed.
  • Thirteen participants with a mean age of 42 years and a mean A1c of 7.4% took part in the study. Of the 13 patients, 11 were female. Mean BMI was 24.7 kg/m2, mean duration of diabetes was 27.2 years, mean duration of pump use was 9.6 years, and insulin brands included seven patients on Humalog, five patients on Novolog, and one patient on Apidra.
  • The HHM kept participants in target range (70-180 mg/dl) nearly 70% of the time and achieved a mean blood glucose of 165 mg/dl (standard deviation: 39 mg/dl) (statistics as measured by YSI). As expected, time in range rose significantly over night (over 80%) and was most challenging after breakfast (55%). We were glad to see time in range metrics reported for both CGM and YSI values – this is not often done in AP studies but is useful from a CGM accuracy perspective. Additionally, we note that the two analysis methods were largely congruent except in the post-lunch period (see table below). Drs. Aaron Kowalski and Boris Kovatchev have often said that current CGMs are accurate enough for control to range, a sentiment that is supported by this study. Overnight was defined as midnight to 7:00 am, post- breakfast was 7:00 am to 1:00 pm, and post-lunch was 1:00 pm to 8:00 pm.

Mean Percentage of Time Spent at Different Glucose Levels Based on CGM and YSI Measurements

 

YSI

CGM

Overall 70-180 mg/dl

69.6

62.2

Overall <70 mg/dl

0.2

0.5

Overall >180 mg/dl

30.2

37.2

     

Overnight 70-180 mg/dl

81.8

80.0

Overnight <70 mg/dl

0.3

0.1

Overnight >180 mg/dl

17.9

19.9

     

Post-breakfast 70-180 mg/dl

55.3

53.4

Post-breakfast <70 mg/dl

0.3

1.5

Post-breakfast >180 mg/dl

44.4

45.1

     

Post-lunch 70-180 mg/dl

63.3

46.9

Post-lunch <70 mg/dl

0

0

Post-lunch >180 mg/dl 36.7 53.1

PERFORMANCE METRICS OF A HYPOGLYCEMIA–HYPERGLYCEMIA MINIMIZER (HHM) SYSTEM IN A CLOSED-LOOP FEASIBILITY STUDY (922-P)

Ramakrishna Venugopalan, Daniel A. Finan, Thomas W. McCann Jr., Linda Mackowiak, Eyal Dassau, Stephen D. Patek, Henry Anhalt

This follow-up poster to the study described above details the performance and behavior of the control algorithm used in the 13-patient, in-clinic study of Animas’ hypoglycemia-hyperglycemia minimizer (HHM) System. The controller appropriately increased and decreased basal infusion to mitigate below- zone and above-zone glucose excursions (it targets a range of 90-140 mg/dl). For below-zone excursions, the system reduced pre-programmed basal rates by 85.7% on average; for above-zone excursions, basal rates were increased by 42.2% on average (as assessed by CGM values). The controller also took preemptive action to avoid below-zone excursions (defined as adjustments to insulin infusion 15 minutes prior to an excursion) 100% of the time when performance was assessed using CGM values and 83% of the time for YSI values. Of course, one of the major challenges with any control-to-range system is the slow speed of current rapid-acting analogs (we’ve heard it characterized as turning the steering wheel on a car and then waiting an hour or more for the car to respond). With this in mind, we especially look forward to faster insulins (e.g., Novo Nordisk’s NN1218, Halozyme’s PH20, Biodel’s BIOD-123, Novo Nordisk’s upcoming decision on the ultra rapid acting insulin to move into phase 3) and faster insulin delivery technologies (e.g., BD’s intradermal needles, InsuLine’s InsuPatch). There is a great deal happening on this front, and we’re very glad since we believe the biggest weaknesses in insulin are by far with prandial vs. basal.

  • The algorithm of the HHM System comprises two components to control glucose to a target zone. A model predictive controller (MPC) uses a mathematical approximation of insulin-glucose dynamics in the participant to predict near-future glucose trends from recent CGM measurements and insulin dosage amounts. The algorithm is designed to deliver insulin as needed with an objective of maintaining glucose levels within the target zone of 90-140 mg/dl. The safety module uses (other) mathematical approximations of insulin-glucose dynamics to continually assess and mitigate the risk of near-future hypoglycemia. It acts on the MPC’s recommended insulin infusion amount, and is designed to provide an additional safeguard against predicted near-future hypoglycemia.
  • As assessed by both CGM and YSI values, the HHM System’s algorithm adjusted insulin infusion to mitigate below-zone and above-zone excursions. Values below are reported as the average of all 13 participants from the feasibility study. The table below excludes all data up to 60 minutes after meals, during which above-zone excursions are anticipated.

 

Ability of the HHM System algorithm to mitigate below-zone and above-zone excursions
  Below-Zone Excursions Above-Zone Excursions
 

Average Basal

Average HHM

% Change from

Average Basal

Average HHM

% Change from

 

Rate

System

Basal

Rate

System

Basal

 

(units/hr)

Rate

Rate

(units/hr)

Rate

Rate

   

(units/hr)

   

(units/hr)

 

CGM

0.9

0.13

-85.7

0.86

1.21

42.2

YSI

0.8

0.27

-67.5

0.87

1.24

46.6

  • The HHM System’s algorithm took preemptive control action nearly 100% of the time when below-zone excursions were predicted. Preemptive action was defined as algorithm adjustments to insulin infusion (relative to basal) in the 15 minutes prior to the start of a below-zone excursion. The table below represents the total number of excursions for all 13 patients in the trial.

Below-zone Prediction Capabilities of the HHM System Algorithm

(number of times preemptive action taken/number of below-zone excursions)

Based on CGM

Based on YSI

9/9 (100%)

5/6 (83.3%)

FEASIBILITY OF ADJACENT INSULIN INFUSION AND GLUCOSE SENSING VIA THE MEDTRONIC COMBO-SET (901-P)

David N. O’Neal, Sumona Adhya, Alicia Jenkins, Gayane Voskanyan, Glenn Ward, John B. Welsh

This feasibility study assessed the performance of the Medtronic Combo-set, which incorporates an insulin infusion catheter and a CGM sensor separated by a short distance (i.e., two skin punctures but a single insertion device). The study showed that insulin pharmacodynamics and CGM accuracy with the Combo-set were comparable to standard, independently located CGM and insulin infusion sites. This is encouraging news and certainly an improvement over current devices, though we hope Medtronic could eventually reduce the Combo-set to just a single skin puncture. (The poster notes that in earlier studies in dogs, the CGM sensor was built in to the infusion catheter wall but it showed interference during both insulin and diluent (placebo) infusion.) We note that Medtronic does not have any ongoing studies of the Combo-set on clinicaltrials.gov and a combined insulin infusion/CGM set was not mentioned in the pipeline discussion at the recent analyst day (see our report at http://bit.ly/LhilPw).

  • The Combo-set incorporates a CGM sensor and insulin infusion catheter separated by a short distance (i.e., two skin punctures but a single insertion device). The poster showed a single inserter for the Combo-set (the “Combo-serter”). An exterior picture of the Combo-set displayed a Medtronic Quick-Set infusion set with a Mini-Link transmitter attached to the top of it. An in situ view showed a skin cutaway version of the Combo-set – the CGM sensor and insulin catheter appeared to be separated by about half an inch.
  • This study aimed to determine (1) if real-time CGM readings are affected by nearby insulin infusion and (2) if insulin pharmacodynamics are affected by a nearby glucose sensor. Ten individuals with type 1 diabetes (mean age: 47 years, mean diabetes duration: 22 years, mean pump use: 6.4 years) participated in the study. Each patient had a Combo-set inserted in the abdomen, a contralateral Sof-sensor attached to an iPro recorder as a control, and a contralateral infusion set for routine insulin delivery. The Combo-set delivered insulin diluent except during meal tests on days one and three, when boluses of insulin lispro were delivered via the Combo-set. Post-bolus venous lispro levels were determined at 0, 30, 60, 120, and 180 min. Capillary blood glucose readings were collected with Bayer Contour Link meters.
  • The accuracy of sensor glucose readings was not affected by nearby insulin infusion. Mean absolute relative deviation was 17% with the Combo-set vs. 18.9% with the Sof-sensor (p=0.63).
 

Combo-set (n=10)

Sof-sensor (n=10)

Mean ARD

17.0%

18.9%

Mean ARD Range

11.5%-36.4%

9.8%-36.9%

Median ARD

13.5%

13.3%

Clarke A + B

96.8%

93.1%

  • Pharmacodynamics of insulin were not affected by a nearby glucose sensor. Insulin via the Combo-set showed the expected post-bolus peak time of 66.6 minutes. Postprandial glycemia after both test meals using the Combo-set was comparable to profiles obtained on day two, when participants were on their usual diet and received insulin via the control infusion set. One "No Delivery" alarm occurred during the 21 patient-days of use, similar to the historical control rate of other infusion sets (1 per 24 patient-days in the CareLink database of 99,857 patients in 2010).

Symposium: Beta Cell Replacement Therapies for Severe Hypoglycemia Unawareness

A BIONIC PANCREAS DELIVERING INSULIN AND MICRO DOSE GLUCAGON AUTOMATES BLOOD GLUCOSE CONTROL IN TYPE 1 DIABETES

Steven Russell, MD, PhD (Massachusetts General Hospital, Boston, MA)

Dr. Russell reviewed the last couple years of bi-hormonal closed-loop work at Mass General. Most intriguing to us was the concluding portion of the presentation, which featured discussion of the team’s upcoming trials in adults and pediatrics. A five-day outpatient closed-loop study is slated to start in late 2012, which may reflect a slight delay from the “summer” timeline we heard at ATTD in February (see pages 18-19 of our report at http://bit.ly/yQtkcM); as of right now, the mobile AP system is still awaiting FDA approval (an iPhone 4S as the controller and two Tandem t:slim insulin pumps to deliver glucagon and insulin), and we certainly hope the Agency keeps the team’s ambitious work moving. What’s noteworthy about this five-day trial is that participants will have free roaming around the Massachusetts General Hospital Campus, no set schedule or diet, and free access to the gym. More ambitious is the planned two-week (!) outpatient closed-loop study at a diabetes camp – this is still slated for summer 2013 (unchanged from ATTD). Although not mentioned today, Dr. Russell and colleagues are also planning a 12-day study of Massachusetts General Hospital staff with type 1 diabetes (2013) and a study in newly diagnosed patients (2013-14). Down the road, Dr. Russell believes the ultimate commercial AP product will ideally have a form factor similar to the current t:slim pump with a second reservoir for glucagon infusion and a built in CGM receiver and controller algorithm (two patch pumps would also be feasible in his view). Also possible would be a sleeve that would fit on the back of a smartphone where the logic and wireless radios would reside. In Dr. Russell’s opinion, the phone would be used only as an interface so that it wouldn't fall under regulation as a medical device (e.g., a similar strategy to Sanofi’s iBGStar). Ultimately, Dr. Russell hopes for a dual infusion set for glucagon and insulin and the potential for the team’s algorithm to work with multiple smartphones, CGMs, and pumps.

  • Dr. Russell emphasized that insulin-only control will not be enough to close the loop. Just one example is exercise, where his team typically sees declines in blood glucose of 2-5 mg/dl per minute. In other words, someone at 70 mg/dl dropping at 2 mg/dl per minute will be hypoglycemic in just 15 minutes. Given this scenario, turning off subcutaneous insulin “will not cut it.”
  • Dr. Russell reviewed previous bi-hormonal studies from Mass General (see pages 18- 19 of our ATTD report at http://www.closeconcerns.com/knowledgebase/r/f0a6108e and pages 57-59 our ADA 2011 report at https://closeconcerns.box.net/shared/dz9hr6m94mu0ehsteybc). He focused on the second study of the system, which used an Abbott Navigator CGM input to the control algorithm, a GlucoScout to measure reference blood glucose, partial meal priming boluses (weight based), structured exercise of around 30 minutes, adults and children ages 12-17 years (no C-peptide), and two control algorithms separately controlling glucagon (PD) and insulin (MPC) delivered through two Insulet OmniPods. The closed-loop system achieved a solid average blood glucose ~158 mg/dl, which correlates to an A1c of 7.1%, 68% time in the range in 70-180 mg/dl (93% at night), and time <70 mg/dl of just 0.7%. The system was also robust to technical failures (e.g., failed OmniPod delivery, computer crash), though the team has asked FDA to modulate some of the algorithms to eliminate even more of the hypoglycemia.

Questions and Answers

Q: I was looking at your blood glucose excursions during meals. You attempted to give more insulin before the meals to reduce excursions. Do you have any comparisons to blood glucose levels during meals in people who don’t have diabetes? How is insulin released before blood glucose starts to rise?

A: We have done a number of experiments. We brought people without diabetes in, had them eat the same meals, and undergo the same monitoring. After a meal of that size, blood glucose may rise to 140- 150 mg/dl in someone without diabetes. You also see some cephalic insulin release before food even begins to be absorbed. The pancreas also dumps insulin into the portal vein, so you see sharp spikes in insulin. We do have a disadvantage giving insulin subcutaneously. I would focus on that fact that we can attain pretty good average blood glucose control and avoid the risk of intravenous and intraperitoneal insulin. As far as we know based on the DCCT, if you achieve a mean blood glucose below 154 mg/dl, there is little signal for microvascular complications. That’s a tremendous improvement over where we are now – cross sectional studies suggest we’re at a mean A1c of ~8.5%.

Symposium: Clinical Aspects of Hypoglycemia in Diabetes – Consequences and Prevention

GLUCAGON AND CLOSED-LOOP INSULIN REPLACEMENT IN DIABETES

W. Kenneth Ward, MD (Oregon Health and Science University, Portland, Oregon)

Dr. Ward provided a thorough scientific overview of stabilizing glucagon for the closed-loop. He especially drilled down into the fibrillation of glucagon, which increasingly occurs as pH is lowered. Dr. Ward emphasized that fibrillation is nearly absent at a pH of 10. This is illustrated in a late-breaking poster [48-LB] from Dr. Ward’s team at this year’s ADA, in which the researchers pumped various glucagon formulations using an OmniPod pump. Glucagon at a pH of 2.5-3 delivered without a pump occlusion for only 47 hours, whereas glucagon at a pH of 10 had not caused occlusion at the 72-hour pump expiration (p = 0.036; n=5). Encouragingly, an in-press study (Ward et al., Clinical Drug Investigation 2012) demonstrates that subcutaneous injection of glucagon at a pH of 10 only led to a slight increase in injection site discomfort (“I don’t think pain in an of itself will prevent us from going forward.”). Dr. Ward believes the most promising glucagon formulation uses a glycine buffer, a stabilizing agent of lactose or bovine albumin, and alkaline preparation at a pH of 9.6-10. Dr. Ward and colleagues also have a number of upcoming studies to better understand and develop a stabilized glucagon in solution. As we noted in our recent report on Xeris (the company is using non-aqueous solvents to stabilize glucagon in solution), the glucagon competitive landscape includes at least four companies using different approaches (for more information, see our report at http://bit.ly/La0rQe).

  • There are a variety of reasons why glucagon fails to prevent hypoglycemia ~25% of the time in the closed-loop setting. First, glucagon fails when there are high concurrent insulin levels. Dr. Ward (with funding from JDRF) us undertaking a study to better understand this. Second, depletion of hepatic glycogen may also be a problem. Studies are needed to develop better models to more accurately anticipate the effects of glucagon. Another cause of glucagon failure is overestimation of glucose by the sensor (not specifically glucagon related, but still a problem). Finally, the chemical instability of glucagon is particularly challenging – glucagon is “not pumpable in its current form” due to polymerization into amyloid fibrils (probably cytotoxic) and degradation of the native glucagon molecule.
  • There are a number of remaining “to-do’s” that Dr. Ward hopes to study going forward: quantify the effect of stabilizing agents on glucagon bioactivity, shelf life stability studies during refrigeration, the biological effect of glucagon in pigs (fresh vs. aged), and the biological effect and tolerability of glucagon in humans (fresh vs. aged).

Product Theaters

INTRODUCING THE T:SLIM INSULIN DELIVERY SYSTEM (SPONSORED BY TANDEM DIABETES)

Kim Blickenstaff (CEO and President, Tandem Diabetes Care, San Diego, CA)

Tandem CEO Kim Blickenstaff led off the t:slim product theater with a major focus on simplicity (for our take on the t:slim approval and experience with the device, please see our November 2011 reports at http://bit.ly/urPT0M and http://bit.ly/Afpuxs). To start, he highlighted a number of well-designed and poorly designed products (well designed: the ATM; poorly designed: BMW’s iDrive, the Metro ticket machines in Washington, DC, and the Baxter Colleague Infusion Pump), explaining that the design process MUST involve the intended users of a device. Mr. Blickenstaff also shared data from the pump’s summative study for FDA approval, in which users typically learned the pump’s interface in about 90 minutes and 70% never referred to the user manual. A narrated t:slim promotional video followed, which was similar to those shown at JPM 2012 (see our report at http://bit.ly/zkvgXG) and have also been posted on Tandem’s website. Highlighted features included the simple user interface, the “slimmest, most compact design,” and “the first touchscreen insulin delivery system cleared by FDA.” Last, Mr. Blickenstaff asserted that the t:slim is easier for educators to teach, easier for patients to learn and use, and may even reduce the support burden placed on diabetes practices. In supporting the last five months of clinical studies, Tandem has also built out its customer support. Tandem announced yesterday that it will begin taking orders for the t:slim on June 11 and the first orders will ship in August 2012.

Jen Block, RN, CDE (Stanford University, Stanford, CA)

The compelling educator-extraordinaire Ms. Jen Block walked audience members through the t:slim’s interface using an on-stage iPad. She first noted how the pump helps her achieve goals as both an educator (simple to train users on) and a patient (less time interacting with diabetes and more time to live life). The majority of her presentation included quick navigation through a variety of activities on the t:slim interface, including the well-designed home screen, the setting up and duplicating of user profiles, and seven-day averages and statistics.

Timothy Bailey, MD (Advanced Metabolic Care and Research, Escondido, CA)

Dr. Bailey concluded the product theater with a summary of t:slim’s three-day home use study prior to FDA approval (n=29) and the ongoing 30-day pre-launch study (n=100). The major takeaway was high patient enthusiasm for the t:slim in both studies. In the pre-launch study, ease of use scores across 14 critical t:slim functions averaged a 6.6 on a seven-point Likert scale (1=impossible to use, 7=very easy to use), well exceeding typical scores for devices. The reception was equally positive in the pre-launch study: 100% of subjects reported being satisfied with t:slim’s size, 95% of patients were satisfied with t:slim’s color screen, and 95% of patients were satisfied with t:slim’s home screen. Patients were also asked to compare the t:slim to their current pump on several metrics: convenience, easy to learn, ease of taking insulin, ability to keep blood glucose stable, and willingness to recommend t:slim to friends. Patients rated t:slim statistically significantly higher than their current pump for all the aforementioned metrics.

PANEL DISCUSSION

Timothy Bailey, MD (Advanced Metabolic Care and Research, Escondido, CA); Jen Block, RN, CDE (Stanford University, Stanford, CA); Kim Blickenstaff (CEO and President, Tandem Diabetes Care, San Diego, CA); Linda Parks, RN, CDE (Tandem Diabetes Care, San Diego, CA)

Q: The pump has a rechargeable battery. Do patients need to sit next to an outlet?

Ms. Block: Some people choose to disconnect, but you don’t have to.

Q: How long does it take?

Ms. Block: It depends on how long you’ve gone. The pump can last seven days on a full charge. We recommend you charge for a few minutes each day.

Q: So the user must sit there with the pump connected to an outlet?

Ms. Block: You could charge it while showering. One thing I’m excited about as an educator is that when you plug t:slim into a computer, it not only charges but it also asks you if you want to upload data. So it could be plugged in, uploading, and charging at the same time.

Q: What about a camping trip? What do you do there?

Ms. Block: It does come with a car charger.

Ms. Park: You can charge in a car, or buy one of those battery packs – it’s just like a cell phone. There are lots of different options.

Q: Is the device waterproof?

Ms. Parks: It’s rated to IPX-7. That means it can be in water for 30 minutes at three feet. We don’t encourage swimming with it, but it’s okay if you’re in the shower or thrown into a pool.

Q: Will this pump have integration with CGM?

Ms. Parks: The first pump will not. But we have an agreement with Dexcom.

Q: Have you seen any issues with the font size?

Ms. Parks: All those crazy things you can do with the iPhone are heavily patented. The font cannot be made bigger.

Q: Were there comments from patients that they couldn’t read it?

Ms. Park: The contrast is bright. We had very few comments on this.

Q: What about customer service issues. Will there be an 800 number available 24/7?

Dr. Bailey: Guaranteed.

Q: Will it be some god awful recorded voice on the other side or a live human being?

Ms. Parks: First you will get a recording to send you to the correct department, and then you will speak to a human.

Q: What about on nights and weekends? Will the lines be staffed by nurses or will it be technicians who have to get a nurse?

Ms. Parks: A mix. We have clinical people and salespeople in the field. We also have nurses and patients with diabetes in customer support.

Q: What is the turnaround time if a present pump doesn’t work.

Dr. Bailey: Fed-Ex the next day. Ms. Parks: Within 24 hours.

Q: I was confused when you were selecting the Gym profile. How does that work?

Ms. Block: It’s activated with the touch. You just go into the options menu and to personal profiles. From there, I can select which one I like.

Dr. Bailey: This also has the cutting and pasting. In pumps, you always had to do everything from scratch. So now you can have a complicated profile and then tweak it.

Ms. Block: And as you’re tweaking it, the profile will appear shadowed.

Q: Can you show a cartridge change?

Ms. Parks: We don’t have a video here. But come by the booth and we can show you how to change it. It’s very easy.

Q: What if the pump has been used for a full seven days and it has zero battery life?

Ms. Parks: With a completely dead battery, it will take one hour to fully charge. It will turn back on in 15 minutes after it’s plugged in. We’re trying to encourage people to top it off every day. For instance, while you’re taking a shower, plug it in and keep the battery charged.

Q: It the touchscreen pressure-based or electrostatic-based?

Ms. Parks: It’s capacitance. You cannot use it with gloves on. Q: How long is the durability of the touchscreen. Ms. Parks: It’s durable and shelf tested.

Q: Can your pump show basal rates with a graph?

Ms. Block: Not on the pump, but it can in the t:connect software.

Q: So in the pump you can only see text?

Ms. Block: Yes.

Q: Can you expand on infusion sets?

Ms. Block: You can connect to any infusion set that has a luer lock.

Q: When you set the Gym profile by duplicating the other profile – what I did not see was when that basal rate stops and the other resumes. You changed it at 5 pm. But that basal rate continued indefinitely to midnight.

Ms. Block: it did and that was my intention.

Q: You said you are approved for ages 12 and over. But you said the participants in your trial were 22 years and older. How did you get approval?

Ms. Parks: In the summative studies for FDA, the youngest child was 12. In the current user evaluation studies, we limited it to 18 years and older. We wanted to test the usability of the pump and prepare for launch.

Q: Can you charge the pump with any USB charger?

A: Any micro USB charger.

Q: Do you have other languages?

Ms. Parks: It’s English only. We have plans to look at other countries after the US launch.

Q: What’s the delivery amount?

Ms. Parks: Every five minutes. Increments up to 0.001 units. The lowest basal is 0.1 units per hour.

Q: How do you calculate insulin on board?

Ms. Parks: Curvilinear.

Q: Is there a carb diary in the pump?

Ms. Parks: There is no carb diary in the pump.

(Editor’s note – the tone of this panel conversation was less than constructive at points and we were surprised and disappointed the audience would ask some openly antagonistic questions.)

Corporate Symposium: Waiting for Closed Loop – Tapping the Full Potential of Advanced Diabetes Technology in Today's Clinical Practice (Sponsored by Medtronic)

INTRODUCTION

Irl Hirsch, MD (University of Washington School of Medicine, Seattle, WA)

Dr. Hirsch opened with a question, “What is the holy grail of type 1 diabetes?” The answer to this question has changed tremendously over time – from improving insulin, to the islet, to the holy grail of today, the artificial pancreas. The latter was the topic of the night’s Medtronic-sponsored symposium. Dr. Hirsch outlined the schedule, which included his presentation on the evidence behind sensor- augmented pump therapy, Dr. Thomas Danne on reducing the risk of hypoglycemia, and Dr. Timothy Jones on the latest progress towards closing the loop.

ADVANCED DIABETES TECHNOLOGIES: THE WEIGHT OF EVIDENCE

Irl Hirsch, MD (University of Washington School of Medicine, Seattle, WA)

In the symposium’s opening presentation, Dr. Hirsch reviewed the path we’ve taken to developing the closed loop. He emphasized the importance of the Star 1 trial (Hirsch et al., Diabetes Technol and Ther 2008), particularly because it revealed how to design a better CGM trial, how to select appropriate patients for CGM, and how to use CGM. Notably, the Star 1 trial also pointed to the relationship between patient adherence to CGM and A1c reduction – a common thread throughout the CGM studies Dr. Hirsch presented and the take home message of his talk. Turning to the JDRF CGM Study (NEJM 2008), he emphasized that we learned our lesson from Star 1. The trial showed a large separation in A1c between adults (>25 years) using real-time CGM and the control group. But again, success was moderated by how often patients wore the sensor. (He quipped to the audience’s amusement, “You can lead a horse to water, but that doesn’t mean he will drink or wear his CGM”.) This relationship surfaced once more in the Star 3 study and data from the Helmsley Charitable Trust’s T1D Exchange. He explained that regulatory officials and payers must realize that success in CGM is dependent on patient behavior, and that a trial using CGM is not comparable to a trial where all you have to do is take a pill. “That’s not the way we do diabetes and that’s not how patients take care of themselves.” Dr. Hirsch concluded his talk by announcing the latest progression in closed loop technology from Medtronic – the MiniMed 530G, which was recently submitted to the FDA and he believes will be available in the US market in 12-18 months (Medtronic management estimated 12 months during the recent Medtronic Analyst Day; see our coverage at http://bit.ly/KiPdWG).

REDUCING THE RISK (AND FEAR) OF SEVERE HYPOGLYCEMIA

Thomas Danne, MD (Kinder- und Jugendkrankenhaus Auf der Bult, Hannover, Germany)

Dr. Danne gave a broad overview of hypoglycemia, emphasizing the critical importance of identifying it in clinical practice, ascertaining its causes, and preventing it. Dr. Danne first reviewed data and professional guidelines suggesting the use of pumps and CGM can help prevent hypoglycemia and severe hypoglycemia. He next discussed Medtronic’s Veo insulin pump with low glucose suspend (LGS), recently submitted to the FDA as the MiniMed 530G (for more information, see the Medtronic 2012 Analyst Day report). We appreciated his review of the multi-center crossover study (LGS on or LGS off) of the Veo published last year (Danne et al., Diabetes Tech Therapeutics 2011). The main finding was that average glucose control did not change with LGS turned on vs. off, while all measures of hypoglycemia improved when patients had the LGS feature turned on (“You’re preventing hypoglycemia and your average glucose control is not getting worse. That’s a lot of good news”). Consistent with previous data, the vast majority of full two-hour LGS episodes occurred at night, the average blood glucose rise during pump suspension was 35 mg/dl/hour, and most alarms occurred during the day and were of short duration. Dr. Danne emphasized that there was no real risk of severe hyperglycemia from a two-hour suspension (one of the FDA’s major concerns with the Veo). In closing, Dr. Danne briefly mentioned the next step, predictive LGS management, which he is “taking a closer look at.” [We learned in the aforementioned Medtronic Analyst Day that the MiniMed 640G, a hypoglycemia minimizer, is expected in early 2013 (EU) and 2014-2015 (US).]

CLOSING IN ON CLOSED-LOOP INSULIN DELIVERY: NEAR AND LONG-TERM IMPERATIVES

Timothy Jones, MD (University of Western Australia, Perth, Australia)

Dr. Jones continued the focus on hypoglycemia in the symposium’s concluding presentation. He asserted that excessive fear of hypoglycemia leads to inappropriate diabetes management: in one survey at his clinic, fear of hypoglycemia caused 13% of children to eat a large snack at bedtime, while 11% of parents often avoided leaving their child alone (age >13 years). In these groups, mean A1c was 0.8% higher. Most unique was his slide showing handwritten responses from children to the question, “What worries you most about hypoglycemia?” Answers shown included “passing out and dying [sic]” “Going so low that I go into a comba [sic] and not waking up,” and “being alone.” Talk about reasons to get an LGS here in the US and predictive LGS as soon as possible... Dr. Jones next reviewed the Australian study of the Medtronic Veo (Ly et al., Diabetes Care 2012), which has demonstrated similar data to the Danne et al. study. The most common problems experienced by patients with the Veo system have been CAL Errors (calibration error between the sensor glucose and meter glucose), weak signal errors, and sensor errors. After six months of using the Veo, 85% of participants elected to continue using it for a second six months; moreover, 94% said they would recommend the system to other patients. While these numbers may reflect a trial effect, it is encouraging to see such high satisfaction with the product and we look forward to its arrival in the US likely sometime in 2013. Dr. Jones concluded with a brief slide showing Medtronic’s portable glucose control system (“the first step to home studies?”), which consists of two MiniLink transmitters with Enlite sensors, a BlackBerry smart phone, a Paradigm Veo insulin pump, a 915 MHZ translator, and a remote monitor (we first saw the system at the 2011 Diabetes Technology Meeting; see page five of our report at http://bit.ly/uh3qXd).

PANEL DISCUSSION

Panelists: Irl Hirsch, MD (University of Washington School of Medicine, Seattle, WA); Thomas Danne, MD (Kinder- und Jugendkrankenhaus Auf der Bult, Hannover, Germany); Timothy Jones, MD (University of Western Australia, Perth, Australia)

Dr. Hirsch: We have had CGM for 12 years. A question for both of you: ‘Where will we be a decade from now?’

Dr. Danne: Pediatricians are always optimists. Ten years from now, we will not talk about closed loop being around the corner. Closed loop will be here. We’ll have a glucose sensor that’s redundant, using two different technologies. It will be so small that you will be able to wear it easily. It will work on its own. It will be a question of whether you want a Porsche or a Mercedes.

Dr. Hirsch: Appropriate from a German colleague [laughter].

Dr. Jones: I think that as human beings, we’re good at technology. Once we start using them, they will improve. Ten years is a long time. I’m confident that it’ll be normal therapy by then.

Dr. Hirsch: How are pumps paid for in Australia and Germany? Please compare and contrast with what happens in the United States with employer-based insurance.

Dr. Danne: We have the lucky situation in Germany that if we put in a good application, the insurance company will pay for the pumps, but we have a difficult time with CGM in Germany.

Dr. Jones: Paying for this is a challenge. Insurers will pay for pumps, but only 50% of people are insured. The National Health system does provide coverage.

Q: Germany has had a lot of international publicity about payment of insulin analogs. Is that still an issue?

Dr. Danne: Payment is always an issue. Yes, we have payment for rapid-acting analogs. But in Germany, only 50% of type 1 diabetes patients are actually using rapid acting analogs. In Germany, most patients are on human insulin. The financing is actually not so much of a problem.

Dr. Hirsch: By far the number one question has to do with reimbursement, not technology. In Australia, is CGM covered?

Dr. Jones: The sensors are not covered at all. They can get it through other means.

Q: Dr. Jones, why do you think that 15% didn’t want to stay on the LGS?

Dr. Jones: They don’t like the alarms, they don’t like wearing something, some are teenagers. It’s the reasons people don’t want to use CGM.

Dr. Hirsch: I actually think that’s a surprisingly low number.

Dr. Hirsch: There are a lot of questions, specifically for pediatricians, about problems with allergies and skin reactions, both in terms of sensors and pumps. What are the tricks you use?

Dr. Danne: I am amazed by the ideas parents come up with. There have been a range of solutions, including adhesives and sprays. I don’t remember a lot of cases where we couldn’t find a solution. It is more size problems and size rotation problems than allergy issues.

Dr. Jones: When we started we saw a lot of those types of problems and it has improved with time.

Q: Is there a validated tool to assess fear of hypoglycemia?

Dr. Danne: There is a fear of hypoglycemia questionnaire. I don’t think it is that sensitive. In my clinical experience, fear of hypoglycemia is much easier to get at with talking to a patient rather than a questionnaire.

Dr. Hirsch: Up to how many times would you let a child correct hypoglycemia with carbohydrate intake? The example you showed did it several times. What do you recommend for that?

Dr. Danne: It depends on patient tolerance. I don't have too many patients who pull out their pump when they have hypoglycemia. Most correct with carbohydrates. This was an extreme patient who had had many of these alarms. Most patients who are not on a sensor will do it by clinical symptoms.

Q: How did you define hypoglycemia unawareness? Did you see any differences in hyperglycemia based on duration of suspend time?

Dr. Jones: We used the Clarke’s questionnaire. You can use the Clarke’s or Gold’s questionnaire. The Clarke’s is more complicated. There was no relationship in terms of hyperglycemia.

Dr. Hirsch: Have either of you observed differences in LGS in relation to proximity of prior insulin bolus? Do you counsel patients with LGS systems to behave differently in these conditions after bolus?

Dr. Danne: Patient intervention is the rate-limiting step in all types of closed loop systems. The problem we have in the closed loop approach is if you over bolus, you can have the best closed-loop system in the world but it still can’t pull insulin out of your body. Patients who don’t interact with system have the best management. If you severely over bolus, there isn’t much you can do.

Q: When are the best times to calibrate CGM? Pre-prandial, after fasting, during a plateau? Is it best to calibrate during hypoglycemia?

Dr. Hirsch: The way I understand it is to calibrate when the glucose is relatively flat. Dr. Francine Kaufmann (Medtronic Diabetes, Northridge, CA): See the Enlite poster.

Dr. Hirsch: See the late-breaking Enlite poster. I should mention Enlite is not approved in the US. Fran, what do you recommend with your current sensor?

Dr. Kaufman: Stable glucose with a three to four times per day calibration. Dr. Hirsch: That’s what we recommend at our hospital.

Dr. Jones: We recommend calibrating at a stable time between meals. Four times per day.

Dr. Hirsch: Dr. Jones you found two-hour suspend events on 10% of the nights. Since 50% of patients respond to the alarms, does that mean 20% of patients would have two-hour suspend events on any night?

Dr. Jones: 143 patients did not respond to the alarms and had a full two-hour suspend.

Dr. Hirsch: What’s so interesting is so many of these questions deal with reimbursement.   It seems like we don’t have to convince people that the technology is good and it works. The convincing part of it is with the payers.

Dr. Jones: It doesn’t matter who the payers are, they’re reluctant to pay.

Dr. Danne: We also need to be making sensors convenient enough to wear on a continuous basis. We do have cases where there is reimbursement and the sensors are sitting on a shelf at home.

Dr. Hirsch: This is for the pediatricians with regards to very young kids and the topic of real estate. Where on the body can you use this technology for insulin delivery or sensors?

Dr. Danne: I am amazed with what people come up with. The buttocks is excellent, while the abdomen generally is something the kids hate. The thigh works and the arm works. There are many places that are not the official way that still work, so wherever you have subcutaneous fat, try it.

Q: If sensor-augmented pump therapy is state of the art, but only at most half of patients are willing to participate, shouldn’t we investigate more about diabetes coping strategies? How did you address that in STAR-1?

Dr. Hirsch: We did not do well in those people. Those are not the best patients to use this technology. If they don’t use the CGM, they won’t improve. That’s a critical point. One of the conclusions of this session is that no matter how good the technology is, if you don’t have patients willing to participate and use the therapy, it doesn’t matter that much. I don’t know the best word – motivated is one. You have to have a willingness to participate. The original question is about better understanding how to get patients to be more interested in managing their own diabetes. This is a huge, huge issue…

Dr. Kaufman: It’s also with type 2s. They don’t even take their pills.

Dr. Hirsch: It’s with any medication. Look at HIV patients who don’t take their medication. Transplant patients that don’t take their meds.

Dr. Danne: If you had a car that alarmed all the time, that hurt to drive because you were inserting two catheters all the time, that gave you alarms when you have misbehaved, you would probably say to your wife after half a week, “I’ll take the bus.”

Dr. Hirsch: I would turn off the alarms.

Dr. Danne: I really don’t think it’s motivational. It’s a burden on the patient right now. It’s too much. I wouldn’t drive a car either. I would also take a bus.

Dr. Hirsch: Let’s go back to the beginning of insulin therapy. Since then, we have increased the burden by making diabetes more complex. Back even thirty years ago, the majority of patients were on once or twice daily insulin. They took insulin and forgot about it – there was no SMBG. Nothing happened until the eye doctor looked in their eye and said “We have problem.” Or there was edema showing. Until that happened, there was no burden at all. We’ve added burden and it’s come at a great cost, but when you look at the data in the T1D Exchange, the population-based risk of complications has gone down.

Comment: I would disagree. It was still a burden back then. I think back to the patients with type 1 who stood in front of our group when I was a medical student. They said, “The doctor says we should have a normal life.” But they had this huge emotional burden to testing their urine. What you had to do every day back then was not as many steps. But I would disagree that it was not burdensome.

Dr. Hirsch: How many patients did their urine testing?

Comment: These patients did.

Dr. Hirsch: Well, most did not. Most people are not very successful. Look at A1cs in the T1D Exchange. On the other hand, we’re not seeing the complications.

Dr. Danne: Absolutely, Irl – we have seen progress. I’ve saw complications in young adults when I started. I hardly see them now. But we’re still not there, and it’s not right to blame the patient. We haven’t made the advances to be good enough at this point. We’re on the verge of a breakthrough with the closed loop. Five years from now, we might look back and say what we did in 2012 was Stone Age.

Private Event: JDRF/NIDDK Closed-Loop Control Research Meeting

This special meeting has been a fixture at ADA for a few years, starting with only 10-15 people in 2007. Today, the meeting was another who’s-who of diabetes, filling an entire hotel ballroom with members of academia, industry, funders, and FDA. Nineteen separate groups in the US, Europe, Middle East, and Australia are now being funded by JDRF, NIH, or the AP@Home project (EU only). The format for today’s meeting was a panel discussion with six key principal investigators followed by a ‘science fair’ with hands on demonstrations of the technology being used by 12 of the groups. Wow!

OUTPATIENT CLOSED-LOOP STUDY PROGRESS AND PANEL DISCUSSION

Panelists: Stuart Weinzimer, MD (Yale University, New Haven, CT); Roman Hovorka, PhD (University of Cambridge, UK); Boris Kovatchev, PhD (University of Virginia, Charlottesville, VA); Bruce Buckingham, MD (Stanford University, Stanford, CA); Ed Damiano, PhD (Massachusetts General Hospital, Boston, MA); Ken Ward (Oregon Health and Science University, Portland, OR)

Dr. Weinzimer: Before we begin the panel discussion, everyone will give an orientation explaining what they’re doing.

  • Roman Hovorka, PhD (University of Cambridge, UK). The Cambridge group has been developing the “FlorenceD” home closed-loop prototype. The system consists of the Abbott Navigator (currently “the most accurate CGM available in Europe”), the Companion (a device to assist communication with the Navigator), a small laptop running an MPC algorithm, and a Dana R Diabecare pump (with a Bluetooth connection). The switch to the Diabecare pump is a departure from presentations in the past, where the team always seemed committed to the Abbott Aviator pump. Since May, the Cambridge group has been approved to conduct overnight outpatient trials by the UK regulatory authority (MHRA). A typical trial design is a crossover study comparing several nights of open and closed-loop periods. For the first few closed-loop nights, a nurse stays nearby in case of problems. Dr. Hovorka wishes for more robust wireless connectivity, access to the communication protocols, and more pump functionality such as a short basal infusion time and storage of bolus wizards.
  • Boris Kovatchev, PhD (University of Virginia, Charlottesville, VA): Dr. Kovatchev reviewed the iAP study group’s (UCSB/Sansum, Montpellier, Padova/Pavia, and the University of Virginia) DiAs system that we first saw at DTM 2011 and again at ATTD 2012 (see our report at http://www.closeconcerns.com/knowledgebase/r/f0a6108e). The system includes a Dexcom Seven Plus, an Insulet OmniPod with an iDex PDM, and two cell phones (one to run the control algorithm and another to transfer the signal to the iDex). The system has been used successfully used in an outpatient setting in 17 patients with type 1 diabetes for ~700 hours (see our ATTD 2012 report for more detailed accounts). From the first 250 hours of outpatient use, error rates were 3% on the sensor, 2% on the pump, 0.8% with remote monitoring, and only 0.2% with the DiAs running on the cell phone. This led Dr. Kovatchev to conclude, “It does look like a cell phone can run closed loop control.” He concluded with a number of recommendations and thoughts on the future: (1) portable software should be built from the OS level to meet medical device standards; (2) human factor studies are “extremely important” for transition to outpatient use; (3) inter-device communication is still a major problem; (4) algorithm development should include hybrid control; (5) current sensor accuracy is sufficient for control to range. The team plans to use the Dexcom G4 sensor and t:slim insulin pump in upcoming studies.
  • Bruce Buckingham, MD (Stanford University, Stanford, CA). The Stanford group (in conjunction with the University of Colorado, Rensselaer, and the Jaeb Center) have been working on a number of initiatives, including (1) in-home predictive low glucose suspend using a Kalman filter and a Medtronic pump and sensor; (2) inpatient metabolic control at the onset of diabetes using a Medtronic PID controller; (3) control to range studies with an MPC controller and theSansum APS platform; (4) treat-to-range studies with a modified ePID controller; (5) safety studies to detect infusion site and sensor failure; and (6) investigating the University of Virginia’s DiAs system (with remote monitoring) at a children’s diabetes camp. The (1) predictive low glucose suspend work has FDA approval for 1,600 in-home nights with 44 patients, and has already amassed 250 nights of data with 20 subjects. Patients sleep with a laptop at the their bedside. This system has already halved the number of nights at which patients go below 60 mg/dl. Dr. Buckingham wished for more robust wireless communications and a single, simple handheld device with a minimal number of alarms.
  • Ed Damiano, PhD (Massachusetts General Hospital, Boston, MA). Dr. Damiano provided an overview of the Boston group’s bi-hormonal closed-loop research, starting with animal studies in 2006 and followed by the first experiments in humans in 2008-2009 using venous blood glucose as the controller input. From 2010-2012 (>2500 hours of control), the team has used the Abbot Navigator CGM as the input to a laptop-driven controller – studies have included six carb-heavy meals and a meal priming bolus. Dr. Damiano then reviewed the three outpatient studies planned for the next year. All will use either the Abbott Navigator CGM or Dexcom G4, an iPhone 4s controller with an algorithm running in C++, and two tandem t:slim insulin pumps to dose glucagon and insulin. Dr. Damiano showed the room the Navigator/iPhone handheld, which is enclosed in one unit but includes the Navigator receiver hardwired to the iPhone through the 30-pin connector. The system uses low-energy Bluetooth to avoid too much battery usage. This system will be submitted to FDA for an IDE in the next six to eight weeks, while Dr. Damiano hopes to build a G4 version of the system by the end of the year. He hopes to start the transitional five-day study on the MGH campus this fall and finish in fall 2013. A study is also planned for summer 2013 in campers at Clara Barton/Camp Joslin and for 2013-2014 in MGH staff.
  • Ken Ward (Oregon Health and Science University, Portland, OR): The Oregon group has been focusing on the issue of sensor redundancy to eliminate ‘large sensor errors’ (“egregious” errors of over 50%). Studies show that a single Dexcom Seven Plus sensor chosen at random has on average 19 hours per month of large errors. Using two sensors can reduce this to seven hours per month. While this is a lot better, it is still not perfect and can cause problems. The group is also performing inpatient studies with two Dexcom Seven Plus sensors, two Insulet OmniPod pumps (for insulin and glucagon) and a tablet computer. The group is testing the system for eventual outpatient use, and improving the user interface – they expect to move to a Motorola ES400 smartphone and a special belt. The team is also trying to stabilize glucagon for use in bi- hormonal systems and has had success using reducing sugars to dramatically reduce the breakdown (see Dr. Ward’s presentation on glucagon from Day #1 of ADA 2012 at http://www.closeconcerns.com/knowledgebase/r/e83c9c63). Dr. Ward’s #1 wish list item is a single site device that combines insulin and glucagon infusion on the same catheter with sensors on the outside surface.

Dr. Weinzimer: As we move towards outpatient studies, it sounds like some are sticking with one sensor and some are using multiple sensors.

Dr. Ward: That’s an important question Ed and I have been talking about. As sensors get more accurate, it’s quite possible that the value of redundancy declines. But the question is will you still be able to keep the risk of large errors to a minimum. We’ll continue to look at this.

Dr. Damiano: Yes, it’s about the improvement in MARD that’s to be gained from multiple sensors.

Dr. Hovorka: We presented at this year’s ADA that the Navigator had no over-reading of 40% or higher lasting more than one hour. We believe one sensor is sufficiently safe .

Q: This is a fantastic discussion. In the studies from the consortium, we’re seeing great efficacy. But where the rubber hits the road is in system failure and reporting of system failure. Ed, you and I have talked about this. When we clump together patients and present means and averages, it hides what we’re all concerned about: outliers when there are system failures. It’s something to consider as a community. As we move to outpatient studies, how frequently are we defaulting to basal rates? Why are they defaulting? It would help funders, FDA, and industry. To start to narrow down and clump together errors.

Dr. Weinzimer: Maybe there should be agreed upon common reporting metrics.

Dr. Kovatchev: Aaron, what you’re saying is extremely important. There was a quick table in my presentation that showed the frequency of errors for various components. That may be just the first approach to defining what it means for the system to fail. In addition to by component, the duration, and gravity of the failure should also be reported.

Dr. Hovorka: In our studies, we focus on efficacy, safety, and utility. Utility is how long the system is operational. This was our attempt to capture this important issue. I agree, it should be captured.

Dr. Kowalski: Somebody brought up the communications. The diabetes community has struggled with a variety of different ways to download devices. We’ve all seen the pictures at clinics with many cables to download. JDRF is funding a project in Toronto, Canada to get the ball rolling on communication standards that can be used. We’ve reached out to industry. I would urge all of you to introduce yourselves to Joe. The Canadian clinical trials network is working on this.

Q: I wear a CGM and an insulin pump and I have a love-hate relationship with my Dexcom. Have you tried to incorporate other types of data? What kinds of things might trigger failure in the sensor? Heart rate monitors, activity monitors, dietary data, schedules, calendars, would those types of things help inform sensor accuracy?

Dr. Kovatchev: The short answer is all of the above. First, the system itself as its running now doesn’t rely on the sensor. There are two sources of information – one is the sensor and one is the insulin pump. The output of the insulin pump can actually flag sensor errors. Mark Breton is incorporating heart rate in closed loop funded by NIH. There are also studies looking at meal profiles and behavior profiles to flag events daily and take patterns into consideration. That will help when considering sensor accuracy. All that you mentioned is under consideration.

Bruce: If you went to the exhibit, there are patch pumps that have accelerometers built into them. Anything stuck on the body can easily do that. Where I think we need to go is if a cell phone can do face recognition, it should do food recognition.

Dr. Ward: Sweating, showering, jarring the device, those will cause rapid rises and rapid falls in the signal. The big problem is the slow drift you see. We don’t understand the cause of that drift.

Q: Each group has done phenomenal work. Now that we’re moving to the outpatient setting, studies are bigger, more expensive, and more complicated. Could you do joint clinical studies that share economies of scale. Using multiple different camp sites and even a shared control group? Would regulatory bodies consider that?

Dr. Kovatchev: I reported on behalf of a group of four different centers.

Dr. Hovorka: The European commission project has six centers that work together.

Dr. Damiano: I second that. A lot of our initial studies were in CRCs. CRCs can only allow us to do so much. Once out of the CRC, collaboration really becomes the obvious thing to do. Parallel studies and running at multiple sites. We intend to do that.

Dr. Buckingham: The JDRF consortium has wanted that and we should be doing that and sharing the data. A shared control group make sense. Also sharing set failure data.

Dr. Weinzimer: We’re still at transitional studies and not to the point where we can agree on something. Things are still in a weeding out direction and we’re still casting a wide net.

Dr. Weinzimer: Another question on human factors. Have any groups gone out and done some focus grouping activities to look for what users want out of a device? I imagine that many adults with type 1 diabetes may have different expectations than an adolescent or child.

Dr. Kovatchev: In preparation for our outpatient studies, we went through a rigorous cycle of focus groups and human factors. We cannot give that device to a patient without these types of studies.

Dr. Ward: We’re doing these to establish clarity. When you press a button, what does that mean?

Dr. Hovorka: We did an initial focus group two years ago. We’ve haven’t done formal human factors work but it’s on our list.

Dr. Damiano: We’ve been building out the interface for the device I just showed you. In our five-day study, we hope to include a human factors study as part of it. There will be good oversight and one to one nursing. We can run in that a human factors study. Up until now, subjects have not interacted with the device. They could look at it at their bedside, but it’s the first time where they’re interacting with mobile platforms. We’ll learn a great deal over the next 12 months.

Dr. Buckingham: That’s one of the things FDA wants from us. We’re all doing it. I as a physician am also a beta tester for engineers. We’re a hard group.

Comment: I’d like to echo Aaron’s comment. I come from the auto-antibody field, and progress depended on everybody getting together and saying you cannot publish a paper unless you do these things. The group needs to get together and standardize. It really helped in antibodies.

Comment (from FDA): I’d like to make a comment about egregious errors. As you people submit data to us, it’s harder to get data for errors as studies go to the outpatient settings. There is no reference standard. Is there any way to analyze severe CGM errors and see what was the impact of that? That would be extremely helpful for us to understand the risks and how to mitigate them.

Dr. Ward: A number of us are interested in studying closed-loop devices using systems that are not yet approved. The Dexcom G4, Enlite, etc. Is there a special request you have for bringing IDEs for devices that are not yet approved.

Dr. Courtney Lias (FDA, Silver Spring, MD): We definitely encourage the study of new devices. That’s where things are going. As far as outpatient studies, maybe we need more information. You don’t need to have approved devices to get an IDE. You just need to show that there are mitigation factors in case of malfunction.

Q: How much effort is being put forth into creating a system that is more cross platform? Rather than having a patient with their regular phone plus a phone for closed loop.

Dr. Kovatchev: Our control algorithm is written in JAVA. That conversation is long and we should have that outside.

Dr. Hovorka: I would mention the project running in Canada. It’s trying to find a common means to communication. That cross platform connectivity is so essential.

Comment: Regarding the standards, lots of work has been done by industry. There are drafts for insulin pumps and CGM. The intent is to accelerate those projects.

Dr. Kowalski: Before everyone gets out of the room, I’d like to leave a JDRF thought. It stems from data that are very important – the Helmsley Charitable Trust data. It’s very, very important to say that there is a significant unmet medical need. In the Helmsley data, average A1cs are above 8.5%. Then you look at the prevalence of severe hypoglycemia in adults with type 1 diabetes. We still have a tremendous amount of work to do. So many in the audience are affected by type 1 diabetes. There are challenges here. But there is a huge, huge potential to help many, many people.

-- by Adam Brown, Hannah Deming, Kira Maker, Nina Ran, Lisa Rotenstein, Joseph Shivers, Tanayott Thaweethai, Alasdair Wilkins, David Zhang, and John and Kelly Close

 


1 As a reminder, if a patient failed to drop to 85 mg/dl within 10‐15 minutes of exercise, he or she would take a break before re-­initiating exercise up to five more times in the visit. Also, if a patient’s YSI blood glucose value ever fell below 50 mg/dl or rose above 300 mg/dl, the experiment was immediately stopped and treatment administered. At ATTD 2012 Dr. Garg said that due to these various requirements, roughly one-­ third of the inductions had to be repeated.